vu/TODO.md

# Open work

A real backlog, not scattered code comments. Update this file when something
here gets done, and add to it when new gaps turn up — don't let it drift back
into "only findable by reading commit messages."

## Priority

Work in this file follows one methodology, repeated every time a new hour or content type
gets built: **mechanism first, then a minimal live-verified proof (1-2 clean dates/entries
authored fully, end-to-end) to confirm the mechanism is right, then bulk content
authoring/import against the whole calendar once it's trusted.** A section marked "mechanism
built" but not yet "full sweep" or "all N done" is mid-pipeline, not stalled — its next work
is more content in the same already-proven shape, not a design change. This file's dated
build log (below "Open work, in priority order") is the record of that pipeline running
across every hour and content type in this app.

## Open work, in priority order

1. **Mechanism gaps** — a real design/code piece not built yet, blocking any content pass on
   it from starting:
   - (none currently open — the last one, Paschaltide's "Suffragium Paschale" substitution,
     was closed 2026-09-04, see the dated log below)

2. **Minimal-proof-done, bulk content pending** — mechanism is built and trusted (a proof
   date/entry already resolves correctly end-to-end); what's left is authoring/importing more
   content in that same shape across the rest of the calendar:
   - (none currently open — Lauds/Vespers duplex-majus+ Common-category psalmody reached full
     10/10 both hours with `common-of-a-dedication`'s Lauds file, closed 2026-09-05; Matins
     Nocturn 3's last 6 unauthored files closed the same day — see the dated log below)

3. **New features (not content mechanism)** — genuinely new product surface, not something a
   content pass is blocked on; each needs its own design pass, not just implementation:
   - Date picker: `day-nav.ts` currently only steps a day at a time (prev/next); there's no
     way to jump to an arbitrary date without walking there one day at a time or editing the
     URL. Needs a calendar-grid picker UI — open, not started.
   - Audio ("read it to me") interface: a button per hour/text that plays the office aloud
     instead of requiring reading — e.g. via the Web Speech API (`SpeechSynthesis`), which
     would keep this offline-first (no server-side TTS call, no added deploy dependency) but
     needs bilingual voice selection (La/En) worked out, and quality for Latin pronunciation
     is an open question. Open, not started — needs its own design pass before content/UI work
     begins.

4. **Known, deliberate simplifications** — intentional, not bugs, no plan to revisit unless
   priorities change:
   - `temporal-categories.yml`'s `bySeason` mapping is fully live-verified (2026-09-01 sweep);
     what's still not independently re-verified is the *privileged-period tier* classification
     for a handful of specific stretches: Ash Wednesday and Holy Week are `privileged-feria-
     major` on strong prior knowledge, not directly re-checked; the Vigil of Christmas and
     Pentecost's own Ember days default to that same major tier as a guess pending
     verification; Pentecost's Ember days' actual precedence is flagged in-file as a genuinely
     unverified gap, not a guess dressed up as one (Lent's own Ember days *are* verified).

Everything from here down is the dated build log this priority list is drawn from — kept for
the sourcing/verification detail behind each decision, not itself in priority order.

## Content gaps — mechanism built, data incomplete

### Lauds psalmody overrides + office-bundle overrides — done (2026-08)

Mechanism: `src/hours/lauds-psalmody-overrides.ts`, gated in
`hours/lauds.ts`'s `getPsalmodyOverrideFor`; office bundle
(`lauds-capitulum-/-responsory-/-hymn-/-versicle-<id>.yml` in
`data/propers/common/`) falls back to the plain weekday bundle
independently of whether the psalmody override exists. Fires for a saint
ranked duplex-majus or higher, or a named temporal feast
(`marian-saturday`, `christ-the-king`, `christmas-octave-sunday`).

**Authored — all 28 non-`propers: null` duplex-majus+ saints/temporal
ids, plus the 4 already done going in (`st-lawrence`, `marian-saturday`,
`christ-the-king`, `christmas-octave-sunday`) — full 32/32.** Same
methodology as the P/T/S/N batch: one live `command=prayLaudes` query per
saint's own already-established clean date, Monastic Tridentinum 1617.
Extraction automated this time via a purpose-built HTML parser
(`scratchpad/parse_lauds_html.py`, not committed — a one-off tool, not
application code) after manual grep/sed proved too slow at this volume;
cross-validated against 5 saints' worth of already-known-correct content
before trusting it for the rest, then spot-checked further plus an
automated la==en/HTML-leftover sweep across all 28 outputs. Verified
end-to-end with a runtime smoke test resolving each id's actual
`getLaudsPsalmodyOverride` + all four bundle parts through the real
mechanism, not just checking the YAML files exist.

The 5 antiphons cycle through a fixed pool matching the psalm scheme
every one of these major feasts uses ("Psalmi Dominica"): psalms
92/99/62, the Canticle of the Three Young Men, then the Laudate group
(Ps 148-150, one shared antiphon) — confirmed via the reference engine's
own psalm-number labels, not assumed. This is the *same* 5-antiphon pool
Prime/Terce/Sext/None draw from (Prime=group 1, Terce=2, Sext=3, None=5,
skipping the canticle's own group 4) — real, confirmed structurally (see
St. Scholastica below), not a coincidence.

St. Scholastica needed the same treatment as her P/T/S/N content: no
English exists in the reference engine's data for her Lauds antiphons,
capitulum, versicle, or hymn. Four of her five antiphons and her
capitulum reuse the exact English already translated and verified for
her Terce/Sext/None content (confirmed same Latin, since Lauds and the
minor hours share this pool); the canticle antiphon, versicle, and
6-stanza hymn ("Jam noctis umbræ cóncidunt") are newly translated here.
**Real gap surfaced by this: the Lauds psalmody-override schema
(`hours/lauds.ts`'s `BilingualText`) has no per-language status field the
way `hours/resolve-common.ts`'s `ResolvedText` does** — `psalmParts`
unconditionally wraps every psalmody antiphon in `verifiedText()`, so her
self-translated antiphon English can't carry the "draft" marker
(`.text-draft`) the same content does at Prime/Terce/Sext/None. Flagged
here, not fixed — would need `BilingualText` (or the override schema
generally) to carry a status field through to `psalmParts`/`resolvePsalmody`.

### Prime/Terce/Sext/None minor-hour overrides (2026-08, new mechanism)

Found while investigating why St. Lawrence's octave showed none of his own
content at Prime/Terce/Sext/None either — these four hours had *no*
per-feast override mechanism at all before this, unlike Lauds: the psalm
antiphon always came from the plain weekday default
(`data/hours/{hour}-antiphons.yml`), and the chapter was always a single
fixed per-annum/per-weekday text, regardless of what was actually being
celebrated — even on a saint's own actual feast day, not just during an
octave.

Mechanism: `hours/resolve-common.ts`'s `getMinorHourOverrideId` (same
duplex-majus+ threshold as Lauds' own override, deliberately kept in sync
so it's one shared backlog, not per-hour lists that can drift — see
`resolveOfficeWinner`, which makes this octave-aware the same way Lauds'
already is), `resolveMinorHourAntiphon`, `resolveMinorHourChapter`. Prime
additionally needed `isSundayOrFeastOffice` (its capitulum/Preces choice
between Sunday/feast and ferial forms) made octave-aware — live-verified
Prime's capitulum stays the Sunday/feast form (1 Tim 1:17) throughout an
octave, not just the feast's own actual day, so no separate override file
was needed there, just the existing Sunday-form file reached correctly.

**Authored — all 29 duplex-majus+ saints, done (2026-08).** Antiphon +
chapter for Terce/Sext/None, antiphon only for Prime (its chapter needed
no new file — see above), for every id in the Lauds backlog list above
except the 5 `propers: null`-blocked saints (`st-benedict`,
`all-saints-of-the-benedictine-order`, `st-michael`, `st-maurus-abbot`,
`ss-placid-and-companions`) — those still need a collect authored first.
Two saints (`st-scholastica`, `immaculate-conception`) needed corrections
to the reference engine's own data: Scholastica's minor hours have no
English rendering at all in the source (Matins-only English coverage for
that saint) — translated directly, `status.en: draft` throughout;
Immaculate Conception's antiphons render with diacritics stripped
compared to the word-for-word-parallel Nativity BVM antiphon — restored
accentuation to match, `status.la: draft`.

### Prime/Terce/Sext/None Commons extraction (2026-08)

After the P/T/S/N batch above landed, a systematic diff across all 29
saints' files (not just the pairs noticed in passing during authoring)
turned up real shared-Common groups, extracted into `data/propers/common/
{hour}-antiphon-common-of-*.yml` / `{hour}-capitulum-common-of-*.yml`
files and wired via a new `SaintRecord.minorHoursCommon` field
(`calendar/feasts.ts`) — `resolveMinorHourAntiphon`/`resolveMinorHourChapter`
(`hours/resolve-common.ts`) now fall back to it when a saint has no proper
file of their own for a given hour. Deliberately a *separate* field from
the existing (until now unused) `common` field: live-verified the minor
hours don't always draw from the same Common category as the saint's
overall classification (see the Confessor-Bishop group below) — reusing
`common` directly would have gotten that case wrong.

Four groups extracted, all confirmed byte-identical via `diff` first, none
assumed from category alone:

- **`common-of-apostles`** (all 4 hours, antiphon + chapter): st-matthias,
  st-barnabas, st-james-the-greater, st-thomas-apostle, ss-simon-and-jude,
  st-luke, st-matthew, st-bartholomew — 8 saints, 56 duplicate files
  collapsed to 7 shared ones.
- **`common-of-a-confessor-bishop`**: chair-of-st-peter-at-rome and
  chair-of-st-peter-at-antioch share all 4 hours' antiphons *and*
  chapters; st-martin-of-tours shares only the chapters (his own
  antiphons are proper, about his own life). Real finding, not a
  modeling choice: both Chair feasts are `common: common-of-an-apostle`
  in their own saint record (Peter's rank/collect classification), but
  their P/T/S/N content comes from the Common of a Confessor Bishop
  instead — Peter considered *as bishop* of that particular see, for
  that specific commemoration.
- **`common-of-the-bvm`** (chapter only, Terce/Sext/None): nativity-bvm
  and immaculate-conception share all three; their own antiphons stay
  proper (per-mystery). The Assumption's Sext/None chapters share the
  same *reading body* (Sir 24:15-16, 24:19-20) but close with her own
  proper versicles rather than the Common's plain ones, and her Terce
  chapter is a different reading entirely (Sir 24:11-12, not 24:14) — a
  first-class feast's proper office outranking the plain Common, kept as
  her own fully proper files rather than force-fit into the shared one.
- **`common-of-st-peter`** (None antiphon, all 3 chapters): st-peter-ad-
  vincula and ss-peter-and-paul. Not a real liturgical Common category —
  both feasts are specifically about St. Peter himself, so his own
  proper texts get reused directly. Extracted anyway since the content
  really is identical, just named honestly rather than dressed up as a
  Common of Saints class.

### Paschaltide commons — done (2026-08)

The Mark/John-Before-the-Latin-Gate/Philip-and-James "Alleluia" text
flagged above as an open question is confirmed real, not a coincidence:
the reference engine's own source has explicit Paschaltide variants —
`Commune/C1p.txt`/`C2p.txt`/`C3p.txt` ("Commune Apostolorum/Unius
Martyris Pontificis/Plurimorum Martyrum tempore Paschali") — for exactly
3 of its 12 Common categories: Apostles, a Martyr-Bishop, Several
Martyrs. Confessors/Virgins/Doctors/BVM never get one — Alleluia/victory
imagery is restricted to apostles and martyrs specifically, a real
liturgical distinction, not an app gap.

Mechanism: `resolveMinorHourAntiphon`/`resolveMinorHourChapter`
(`hours/resolve-common.ts`) now try `${hourId}-antiphon-
${minorHoursCommon}-paschaltide` (etc.) first whenever `day.season` is
`eastertide`, `ascensiontide`, or `pentecost` — the same three-season
window as Compline's own Regina Caeli table
(`data/hours/marian-antiphon-by-season.yml`), not just `eastertide`
alone, since "tempore paschali" runs through the Pentecost octave in
both cases. Falls back to the plain (non-Paschaltide) Common when no
`-paschaltide` variant is authored, same silent-fallback convention as
everywhere else.

`common-of-apostles-paschaltide` (7 files) ported directly from
`Commune/C1p.txt`/English `Commune/C1p.txt` — not live-queried — and
cross-checked against st-mark's already-authored, live-queried content
(2026-04-25): identical except the source's own Prime-antiphon asterisk
falls one word earlier than the live-rendered incipit had suggested,
corrected here to match the source. st-mark and st-john-before-the-
latin-gate's own duplicate per-saint files removed (all 7 each);
ss-philip-and-james's chapter files removed (its own antiphons stay
proper — its own Gospel-referencing text, not shared even in
Paschaltide). Only `C1p` (Apostles) ported so far, since no currently-
authored saint sits on a Martyr-Bishop or Several-Martyrs common that
could fall in Paschaltide — `common-of-a-confessor-bishop` and any
future Martyr-common saint don't need one (Confessors never get a
Paschaltide variant at all; a Martyr one would need porting `C2p`/`C3p`
the same way, if and when a saint needing it is authored).

### 94 `propers: null` saints — collects authored (2026-08)

**The reference engine's per-date `Sancti/`/`SanctiM/MM-DD.txt` files are
era-merged, not era-specific — blind extraction from them is unreliable.**
Live-querying (`version=Monastic Tridentinum 1617`) always resolved this
correctly elsewhere in this project because the *engine itself* applies
the right rubric filtering; reading a raw `Sancti/MM-DD.txt` file
directly bypasses that filtering entirely. Those files accumulate
centuries of saints assigned to the same calendar date across different
eras, gated behind `(rubrica ...)`-tagged conditional blocks — reading
the file's own untagged/default content silently returns whichever saint
is *currently* primary there (often a post-1617, sometimes post-1888
addition), not necessarily the one relevant to this app's target era.
Confirmed the hard way: an initial blind pull returned St. Didacus's
collect for All Saints of the Benedictine Order, St. Rose of Lima's for
Ss. Felix and Adauctus, St. Josaphat's for St. Martin I, and 9 more
wrong matches out of 13 "clean, unambiguous" test cases — all real
saints, just centuries later than 1617, and none of them what this
app's own calendar actually intends for that date.

**The fix**: `web/www/Tabulae/Kalendaria/` has real, era-specific
calendar *tables* (not per-date files) — `1570.txt` (Tridentine base)
and `M1617.txt` (Monastic 1617's own diffs against it) — giving a
direct, unambiguous answer for "which saint(s), what rank, commemorated
with whom" on every date this app cares about, plus a `fileref` field
(e.g. `01-15t`, `08-23o`, sometimes pointing at a wholly different
MM-DD) identifying *exactly* which source file variant actually holds
that entry's content — including cross-file (`ex Sancti/05-08`,
`Commune/C1v:Oratio 2 loco`) and whole-file (`@Sancti/05-08` as the
file's own first line, no `[Section]` around it) inheritance, all
resolved by a purpose-built parser (`scratchpad/parse_kalendaria.py` +
`extract_final.py`, not committed — see the P/T/S/N and Lauds batches'
own parser scripts for the established one-off-tool convention).

Cross-referencing all ~205 saints in `data/calendar/saints/*.yml`
against this table (not just the 94) found the existing calendar data
essentially sound — 204/205 confirmed correctly dated (only the
deliberate `example-confessor` placeholder didn't match, as expected);
the earlier wrong-saint problem was specific to *this* extraction
method, not a systemic issue with the app's own calendar. No corrections
to `sanctoral-calendar.yml` were needed.

**Resolved St. Michael's own long-flagged uncertainty** (`st-michael.yml`
used to carry a "FLAGGED" comment about Sep 29 showing nothing live
across 6 tested years): the table confirms Sep 29 genuinely is his day
under Monastic 1617, rank 3 — it just never won outright in any tested
year. His own source file is a whole-file cross-reference to the
Apparition of St. Michael's (`ex Sancti/05-08`), confirmed word-for-word
identical live — `propers` now points directly at
`apparition-of-st-michael`'s own id rather than duplicating content.

**Two collect shapes, both now complete for all 94:**
- **34 have a genuine proper collect** — Latin pulled directly from
  source (`{id}-collect.yml` in `data/propers/common/`), English
  translated and marked `status.en: draft` for the ~11 where no source
  rendering exists at all (same convention as St. Scholastica's content
  earlier this session), verified/reused where the source does have one.
- **60 draw from a Common template** — a real architectural addition,
  not a workaround: `SaintRecord.collectCommon`/`collectName`
  (`calendar/feasts.ts`) plus a substitution step in `getDayCollect`
  (`hours/resolve-common.ts`) fill a Common collect's literal `{N}`
  placeholder(s) with the saint's own name(s), positionally — most
  templates need one name, a few (two co-named saints sharing a
  collect, e.g. Ss. Cornelius and Cyprian) need two. 21 distinct
  Common-collect templates now exist (`collect-c*.yml`), named after the
  reference source's own `C`-numbering (`C2`, `C2-1`, `C2a-1`, `C3b`,
  etc.) rather than an invented semantic label, since no clean
  after-the-fact explanation was recoverable for why the tradition
  splits e.g. "Common of a Martyr Bishop" into five differently-worded
  variants. Each saint's own `[Name]` section supplies its declined
  form already in the exact grammatical case that saint's specific
  template needs (confirmed empirically — Ambrose's own `[Name]` is
  accusative, matching his template's own accusative `{N}` slot). One
  real wrinkle: Ss. Popes among the Martyr-Bishop groups (Cletus &
  Marcellinus, Soter & Caius, Anacletus, Anicetus) need the *second*,
  `"(sed communi Summorum Pontificum)"`-tagged declension in their own
  `[Name]` section, not the first — a real, source-documented
  distinction (Popes get their own template variant), not a parsing
  quirk to work around.

One genuine judgment call, flagged rather than silently resolved: St.
Martha's own Monastic-specific source file (`SanctiM/07-29.txt`)
overrides her base Tridentine assignment (`collect-c6a`, one name) with
`collect-c6b` (two names) instead — a plural "Several Virgins and
Martyrs" template that doesn't fit a single saint. Kept the singular
`collect-c6a` assignment as the more sensible reading (see her own
file's comment), treating the Monastic override as likely an
inconsistency in that one file rather than a real joint-commemoration
convention this app doesn't otherwise model.

Verified end-to-end: a runtime smoke test resolving `getDayCollect` for
all 94 ids via a synthetic sanctoral-winner day, confirming a real
(non-`missing`, no leftover `{N}`) collect for every one — not just that
the YAML files parse.

### 94 `propers: null` saints — Benedictus antiphon + P/T/S/N Commons (2026-08)

The natural next step the collects batch unblocked. Two more real gaps
found and fixed along the way, both explained below.

**Raw `Commune/*.txt` fields don't reliably match what the engine
actually renders per hour, same lesson as the collects pass but for
antiphons this time.** `Commune/C2.txt`'s (Martyr-Bishop) own `[Ant 1]`/
`[Ant 2]`/`[Ant 3]` fields don't correspond 1:1 to Prime/Terce/Sext/None
the way a naive reading would suggest, and no `CommuneM/*.txt` file
defines explicit `[Ant Prima]`/`[Ant Tertia]`/`[Ant Sexta]`/`[Ant Nona]`
fields for any of the categories this batch needed (only `C10`, BVM
Saturday, has them). Confirmed independently: the already-authored
`common-of-a-confessor-bishop` P/T/S/N content (from the original
29-saint pass) doesn't match `Commune/C4.txt`'s own raw `[Ant 1]` either
— it was live-query-sourced from Chair of St. Peter/St. Martin's own
rendered pages, not read from the Commune file directly, which is
exactly why. Fix: queried `web/cgi-bin/horas/officium.pl` directly
(local Perl, not a network fetch — `prayPrima`/`prayTertia`/`praySexta`/
`prayNona`/`prayLaudes`, `version=Monastic Tridentinum 1617`) for one
representative `propers: null` saint per Common category, parsing the
antiphon (incipit + full, reconstructing the `*` split point) and
chapter (citation, text, closing versicles) out of the rendered HTML —
new scratchpad tool `query_hours.py`/`batch_hours.py`, not committed,
same one-off-tool convention as the rest of this project's extraction
scripts. Caught its own version of the collects batch's "wrong saint"
bug this way too: St. Emerentiana's stored clean-year comment
(2029-01-23) actually rendered St. Hilary of Poitiers's office by that
year (a stale date after his own earlier Jan 23 → Jan 14 correction,
see above) — every representative's title is now verified against an
accent-insensitive name-stem match before its content is trusted, with a
fallback year search when the stored date doesn't check out.

**The real liturgical categories are coarser than `SaintRecord.common`'s
naming suggests — verified by cross-querying multiple saints per
category, not assumed.** `common-of-a-martyr-bishop` and
`common-of-a-martyr` render byte-identical P/T/S/N and Benedictus
antiphon content (confirmed: St. Anacletus, St. Simeon, St. Blaise, St.
Pantaleon — four different collect-template variants, one shared minor-
hour Common); so do `common-of-an-abbot`, `common-of-a-confessor-not-
bishop`, and `common-of-a-confessor` (St. Odo, St. John Gualbert, St.
Nicholas of Tolentino, St. Alexius); so do `common-of-a-virgin-martyr`,
`common-of-a-virgin`, and `common-of-a-holy-woman` (St. Justina, St.
Emerentiana, St. Praxedes, St. Martha). `common-of-several-martyrs`
splits further than its own name suggests, though: plain groups of
martyrs (Ss. Cornelius and Cyprian, Ss. Faustinus and Jovita, Ss. Placid
and Companions — the last a real Duplex II. classis feast, not just a
low-rank stub, confirming the fallback isn't only reached by minor
saints) share one Common, but groups of *Pope*-martyrs (Ss. Cletus and
Marcellinus, Ss. Soter and Caius) render distinct, more triumphal,
allelúja-toned text of their own — a real category the collect
templates already knew about (`collect-c3b`) that the `common` field
alone doesn't surface. One saint, the Dedication of the Basilicas of Ss.
Peter and Paul, turned out *not* to share its `common` field's nominal
category (`common-of-an-apostle`) at all: live-verified its P/T/S/N and
Benedictus antiphon are the Dedication-of-a-Church Common instead (about
the building, matching its own `collect-c8`), not shared with any
apostle content. Net result: 7 new shared Common categories
(`common-of-a-martyr`, `common-of-an-abbot`, `common-of-several-
martyrs`, `common-of-pope-martyrs`, `common-of-a-vigil`, `common-of-a-
virgin`, `common-of-a-dedication`) plus the pre-existing `common-of-a-
confessor-bishop` cover 66 of the 94 saints; the 7th confessor-bishop
saint's own Benedictus antiphon Common file was also new (the original
29-saint pass never needed it — Chair of St. Peter/St. Martin both have
their own proper Lauds antiphon, only their chapters were shared).

**New mechanism**: `SaintRecord.benedictusCommon` (`calendar/feasts.ts`)
— `getBenedictusAntiphon` (`hours/resolve-common.ts`) now falls back to
`benedictus-antiphon-${benedictusCommon}.yml` when a saint has neither a
proper collect-antiphon pair nor... (see below) — deliberately no name-
substitution mechanism built for it (unlike `collectCommon`): every
Common category in current use renders name-free Benedictus-antiphon
text; only the Doctor Common (`C4a`, not used by any saint here) has a
`{N}`-shaped placeholder in the source.

**Real bug found and fixed while wiring this in**: `getMinorHourOverrideId`
gated *all* Prime/Terce/Sext/None lookup — both the per-saint proper file
and the shared-Common fallback — behind the same duplex-majus+ threshold
Lauds' own *psalmody* override uses, on the assumption that only a
strong feast carries its own minor-hour content at all. Live-querying
this batch proved that wrong: a winning Semiduplex saint (St. Anacletus)
still renders its own proper/Common antiphons at Prime/Terce/Sext/None,
not the plain ferial default — the majus+ gate was only ever a proxy for
"this app has authored proper content," not a real liturgical rule, and
became actively wrong once Common-category content started covering
lower ranks too. Fixed by making `getMinorHourOverrideId` eligible for
*any* sanctoral winner; Lauds' own psalmody override
(`hours/lauds.ts`'s `getPsalmodyOverrideFor`) is unaffected and correctly
stays majus+-gated, since that's a real, much bigger per-saint authored
commitment (the whole psalm-antiphon block), not something this batch
adds a Common fallback for.

One saint, Ss. Ursula and Companions, deliberately left without
`minorHoursCommon`/`benedictusCommon`: her collect template
(`collect-c6b`, plural Virgins and Martyrs) has no other member on this
app's calendar to verify a shared minor-hour text against, and she can
never be queried directly herself (a genuinely unwinnable stub, always
commemorated under St. Hilarion) — left honestly unauthored rather than
guessed from the singular `common-of-a-virgin` category, moot in
practice since a commemoration's own minor hours are never rendered.

Verified end-to-end: a runtime smoke test (folded into
`tests/hours/resolve-common.test.ts`) resolving `getBenedictusAntiphon`/
`resolveMinorHourAntiphon`/`resolveMinorHourChapter` for a sample across
categories, including the Duplex II. classis case to confirm the
majus+-gate fix isn't just a low-rank special case.

**Not done in this pass**: the other 26 of the 94 (`propers` now set to
their own id, from the collects batch's proper-collect saints) still
need their own individual Benedictus antiphon authored — same shape as
the original 109-saint pass (own `[Ant 1]`/live-rendered text, not a
shared Common), since a saint with a genuine proper collect usually has
a genuine proper antiphon too. Not yet started.

### The other 26 propers-set saints' Benedictus antiphons — done (2026-08)

Turned out **most of these don't have unique antiphon text either** —
a genuine proper collect doesn't imply a genuine proper antiphon, same
finding as St. Martin of Tours's mixed proper-antiphon/shared-chapter
split from the original 29-saint pass. Live-queried Lauds
(`prayLaudes`) for each of the 26, same title-verified method as the
Common-category batch above.

**15 of the 26 are genuinely unwinnable stubs** (already documented as
such from the collects pass — always commemorated, never the day's own
title, so nothing to query) — same "honestly missing" treatment as St.
Frances of Rome/St. Bibiana in the original 109-saint pass, not
guessed at.

**Of the 11 that do win outright**, 7 turned out to share existing
Common-category text — `benedictusCommon` wired directly to the already-
authored files from the section above, no new content needed: Ss.
Dionysius and Companions, Ss. Felix and Adauctus, Ss. Maurice and
Companions, Ss. Tryphon and Companions (`common-of-several-martyrs`,
despite Dionysius's own `common` field saying `common-of-a-martyr-
bishop` — another case, like Chair of St. Peter earlier, where the two
fields genuinely diverge), St. Chrysogonus and St. Donatus
(`common-of-a-martyr`), the Vigil of St. Lawrence (`common-of-a-vigil`).

The remaining 4 have real, unique proper text — new
`{id}-antiphon.yml` files: All Saints of the Benedictine Order ("Vos
qui reliquístis ómnia..."), St. Agnes (secunda) ("Ecce, quod
concupívi..."), St. Benedict ("Sanctíssime Conféssor Dómini..." — no
English in the source, same gap as St. Scholastica's, translated and
marked `draft`), and the Vigil of St. John the Baptist ("Eréxit nobis
* Dóminus cornu salútis..." — deliberately echoing the Benedictus
canticle's own opening words, fitting for the Vigil of the saint whose
father, Zechariah, first spoke that canticle).

P/T/S/N for these 26 not attempted in this pass — a further layer on
top, same as how the original 109 got Benedictus antiphons long before
their own later, separate P/T/S/N pass (duplex-majus+ subset only). Done
in a later pass, see "`minorHoursCommon` gap swept across the whole
calendar" below — it covers these 26 too, not just the original 109.

### The 15 "unwinnable" stubs' Benedictus antiphons, resolved from source (2026-08)

Direct feedback on the section above: "unwinnable" was describing an
artifact of Divinum Officium's own single-winner precedence model, not
a real content gap — the app doesn't need a saint to *win* a live query
to have real content, since this app already renders commemorated
saints on their own terms (`day.commemorations`) rather than only ever
showing the day's outright winner. The fix: read each of the 15 saints'
own already-validated fileref directly (`extract_final.py`'s
`load_file`/`resolve`, the exact same fileref their collect was already
sourced from during the Kalendaria-table pull — no live query, no
"does this saint ever win" requirement at all) for its `[Ant 1]`
section, falling back to whatever Common its own `[Rule]`/`[Rank]`
"`vide`/`ex CXX`" reference names when there isn't one — the same
proper-then-Common shape `getDayCollect`/`collectCommon` already uses,
just applied to the Benedictus antiphon too.

**14 of the 15 had no `[Ant 1]` of their own at all** — clean Common
fallback, and every one of their `[Rule]`/`[Rank]` C-numbers (`C2a`,
`C2a-1`, `C3`, `C3a`, `C3a-1`, `C4b`, `C6`, `C6-1`, `C6a`, `C7`) falls
inside a family already live-query-verified in the section above, so
`minorHoursCommon`/`benedictusCommon` were wired directly with full
confidence, no new content needed: St. Marcus (`C4b` — a whole-file
`@Commune/C4` inherit, just the title relabeled "of the Popes" — into
`common-of-a-confessor-bishop`); St. Felix (Presbyter), St. Hermes, St.
Saturninus (`C2a`/`C2a-1` into `common-of-a-martyr`); St. Prisca, St.
Thecla, St. Catherine of Siena (`C6`/`C6-1`/`C6a` into
`common-of-a-virgin`); Ss. Machabees, Ss. Marius/Martha/Audifax/Abachum,
Ss. Twelve Brothers, Ss. Alexander/Eventius/Theodulus/Juvenal, Ss.
Felix/Simplicius/Faustinus/Beatrix, Ss. Nabor and Felix
(`C3`/`C3a`/`C3a-1` into `common-of-several-martyrs`). St. Felicitas is
the one genuine new finding: her file points to `C7` ("Commune Unius
non Virginis Martyris"), which isn't a category this project had
touched yet — but `C7`'s own `[Ant 1]` is itself a cross-reference
(`@Commune/C6:Ant 2`) that resolves to text byte-identical to
`common-of-a-virgin`'s own already-verified Benedictus antiphon, so no
new Common file was needed even there.

**The 1 remaining (Ss. Euphemia, Lucy, and Geminian) does have her own
real `[Ant 1]`** — but marked `status.la: draft`, not `verified`, on
principle: a real counterexample turned up in the *same* pass while
double-checking a winnable saint for comparison — St. Alexius's own
file also has a genuine `[Ant 1]` ("Similábo eum..."), yet his actual
live-rendered Benedictus antiphon (already confirmed in the section
above) uses `common-of-an-abbot`'s text instead, not his own `[Ant 1]`
at all. So a raw `[Ant 1]` read is demonstrably not sufficient proof by
itself that it's what the Benedictus canticle actually uses (Vespers'
Magnificat is a live candidate for what it's *actually* for), even for
a saint that *can* be live-checked — and for one that never can be, that
uncertainty can't be resolved at all. Used the text anyway (better than
nothing, and plausibly correct), just honestly flagged rather than
claimed as confirmed.

All 15 now have real content — none left "honestly missing." Ss.
Ursula and Companions (the section above) is the one case that's
actually different in kind: no other calendar member shares her
`C6b`/plural-Virgins-and-Martyrs category to cross-check a Common
against, so even a clean file read wouldn't have anything solid to
verify itself against.

### A full sweep for the same pattern elsewhere, and 2 more real fixes (2026-08)

Same feedback, applied more broadly: swept every saint in
`data/calendar/saints/*.yml` (not just the 94-batch's own lineage) for
a still-`missing` collect or Benedictus antiphon, via a synthetic-winner
smoke test. Found 4 hits; 2 were real, 2 were false positives from the
sweep's own synthetic-day methodology (forcing a saint to be `day.winner`
when in reality it's *never* the winner doesn't correctly exercise its
real content path):

- **St. Frances of Rome and St. Bibiana** (Benedictus antiphon) —
  exactly the fix from the section above, just predating the
  Kalendaria-table work (no fileref had been derived for either yet).
  Both resolve to `common-of-a-virgin` once one is: Bibiana's file
  points to `C6-1` directly; Frances of Rome's points to `C7a` (a
  whole-file inherit of `C7`, just the `[Officium]` title relabeled "of
  a Widow" — fitting, since she's a widow, not a virgin — with no
  override of the same `[Ant 1]` cross-reference, so it resolves
  identically).
- **St. Agatha and St. Lucy** (collect) — a real, older gap, from
  *before* `collectCommon` existed at all: both saints' own file
  comments already documented "no unique collect... resolves honestly
  missing... despite propers being set," written back when that was the
  only option. Agatha's own `[Oratio]` is `@Commune/C6::s/N\./Agathæ/` —
  confirmed identical to the already-existing `collect-c6.yml` template
  — so `collectCommon: collect-c6` was simply wired in. Lucy's own
  `[Oratio]` is `@Commune/C7a`, *unedited* (unlike `collect-c6a.yml`'s
  own C7a-derived text, which has a `"Vírginis tuæ"` insertion baked
  in) — her own `[Name]` section supplies the whole descriptive phrase
  instead ("Lúciæ Vírginis et Mártyris tuæ", not just her bare name),
  making her resolved text effectively proper even though the
  underlying formula is a shared Common one; written as her own
  `st-lucy-collect.yml` (she already had `propers: "st-lucy"` set)
  rather than a new single-use `collect-c7a` template. No English
  `[Oratio]` in the source for either saint's own unedited form, so
  Lucy's stays `status.en: draft`.
- **St. Clare** (false positive) — the sweep's synthetic day forced her
  to be `day.winner`, but she never actually is one (always
  commemorated under St. Lawrence's octave); her real content lives in
  `st-clare-commemoration.yml`, used via a completely different code
  path (`sanctoralCommemorationPart`, not `getBenedictusAntiphon`) that
  the sweep doesn't exercise. Already fully authored and live-verified;
  no gap.
- **Ss. Ursula and Companions** (already covered above) — the one
  genuine, currently-irresolvable case.

Two vigils (Vigil of St. John the Baptist, Vigil of St. Lawrence) didn't
resolve via this file-based method at all (their particular file format
wasn't handled by this pass's parser) — no impact, both already had
real content from the live-query batch above.

### `minorHoursCommon` gap swept across the whole calendar (2026-08)

The pieces above all fixed Benedictus-antiphon/collect gaps. This pass
does the same thing for Prime/Terce/Sext/None specifically, and widens
scope from just the 94-saint batch to the **whole** calendar — the
original 29-saint P/T/S/N pass (an earlier session) only ever covered
duplex-majus+ saints; every below-majus saint from the *original*,
pre-94-batch 109-saint pass had never had P/T/S/N attempted at all,
same gap as the one already found and fixed for the 94-batch's own
below-majus saints.

**Method**: reused the Kalendaria-table-validated fileref approach
(same as the collects pass), not a blind `Sancti/MM-DD.txt` read — a
real near-miss caught this early: St. Apollonia's `Sancti/02-09.txt` is
*currently* St. Cyril of Alexandria's (a 19th-century addition to that
filename slot); blind reading would have silently attributed his
Doctor-of-the-Church Common to her. The Kalendaria table's own
validated fileref (`02-09o`) gives her real content instead. Also fixed
a second, unrelated regex bug the same pass turned up: the original
"`vide`/`ex CXX`" pattern matched the bare word "`Commune`" (no number)
and reading-specific sub-references like "`ex C4 in 2 loco`" (which
only says where *that nocturn's readings* come from, not the overall
antiphon Common) — found via St. Clement, whose Monastic-specific
`[Rule]` override has neither a clean `vide CXX` nor a number after
"Commune." Fixed by requiring a digit right after `C` and skipping any
match immediately followed by `"in N loco"`.

**90 below-majus saints swept, ~82 wired directly** into the categories
already established above (mostly `common-of-a-martyr` and
`common-of-several-martyrs`, the two largest families) — no new content
needed, since every C-number variant found (`C2`/`C2-1`/`C2a`/`C2a-1`/
`C2b`/`C2b-1`/`C2-1b`, `C3`/`C3a`/`C3a-1`, `C5`/`C5-1`/`C5a`/`C5b`,
`C6`/`C6-1`/`C6a`/`C7a`, `C4`/`C4b`, `C1`, `C1v`) already falls inside a
family live-query-verified earlier.

**A real bug this pass caught and fixed**: Ss. Ursula and Companions
(deliberately left `minorHoursCommon`-unwired, see her own section
above) got swept up and *wrongly* wired to `common-of-an-abbot` by this
pass's own script, since she has `collectCommon` set (making her look
like a normal, resolvable batch member) despite `propers: null` and
being a genuine commemoration-only stub — the underlying
Kalendaria-table lookup silently defaulted to St. Hilarion's own
`primary` slot at her shared date rather than her own (nonexistent)
one. Caught by cross-checking every swept saint's `propers` field for
this exact `null`-but-`collectCommon`-set pattern before trusting a
default-to-primary assumption; reverted, she's still correctly unwired.

**A real, new category found: Doctors of the Church are not one thing.**
7 saints' own `[Rule]` says "`vide C4a`" (`Commune Doctoris Pontificis`)
— but live-querying showed this doesn't mean what it looks like it
means. St. Athanasius, St. Augustine, St. Basil the Great, St. Hilary of
Poitiers, St. John Chrysostom, and St. Leo I all render P/T/S/N
byte-identical to `common-of-a-confessor-bishop`'s already-verified
text (their own `C4a` reference just inherits `C4`'s content
underneath, per the same "the referenced file's raw fields don't
reliably predict the live-rendered text" lesson from every other
category this session) — wired directly, no new content. **St. Gregory
the Great is the one genuine exception**: his own live-rendered
P/T/S/N is real, unique proper text (alluding to his own election,
his *Liber Regulæ Pastoralis*, the dove-and-quill legend, and the
"non Angli sed Angeli" episode already known from his Benedictus
antiphon) — 7 new files (`{hour}-antiphon-st-gregory-the-great.yml`,
`{terce,sext,none}-capitulum-st-gregory-the-great.yml`), no
`minorHoursCommon` needed at all (`getMinorHourOverrideId` already
tries his own id first). English wasn't rendered at all for any of his
6 new texts (same Latin-duplicated-into-English gap as St. Scholastica/
St. Benedict), so all 6 are `status.en: draft`.

**A loose thread noticed, then chased down and fixed (2026-08)**: St.
Athanasius (May 2) and St. Leo I (April 11), both tested in years that
happened to fall within Eastertide, rendered their shared
`common-of-a-confessor-bishop` text with a trailing "`allelúja`"
appended to every antiphon — but the other 4 Doctor saints (tested in
non-Paschaltide months) didn't. Live-querying further (a full sweep of
Common categories across the Easter/Ascension/Pentecost octaves and the
ordinary weeks around them, reading the reference source's raw
`Commune/*p.txt` chains directly rather than guessing from rendered
pages) settled it precisely:

- A handful of categories (Apostles, Martyr, Martyr-Bishop, Several
  Martyrs, Pope-Martyrs) have their own dedicated `Cxxp.txt` source
  chain — a genuinely *different* text for Paschaltide, not the same
  text plus a suffix (confirmed by reading the raw files: e.g. the
  "Fíliæ Jerúsalem..." Pope-Martyrs Benedictus antiphon comes from a
  Matins responsory reused for Lauds specifically during this window,
  nowhere near the non-Paschaltide chain's own "Mártyrum chorus..."
  text). Only `common-of-apostles` has this wholesale text actually
  authored into this app's data (`*-paschaltide.yml`, from an earlier
  pass) — the other four still fall through to the mechanical suffix
  below until someone authors their real text, which is a strictly
  better approximation than the unmodified default.
- Every *other* category (Confessor-Bishop/Doctor, Abbot, Virgin, and
  presumably Dedication/Vigil/Angel, untested but structurally the
  same) has no dedicated chain at all — the reference engine just
  mechanically appends a single ", allelúja"/", alleluia" to the
  otherwise-unchanged base text, for the whole Eastertide → Ascensiontide
  → Pentecost-octave-week span. Live-verified across three unrelated
  categories (St. Athanasius/Confessor-Bishop, St. Robert/Abbot, St.
  Catherine of Siena/Virgin) — never doubled.
- The "sometimes 1, sometimes 2" appearance wasn't actually the same
  category flipping — it was different *ranks of saint* getting
  different octave privileges. Only martyr-class saints (already in the
  dedicated-chain group above) are ever eligible for even a
  commemoration inside a true privileged octave (confirmed: St.
  Anicetus, commemorated on a Semiduplex day inside the actual Easter
  octave, correctly showing the doubled dedicated-chain text).
  Confessor/Abbot/Virgin-class saints never get that far — checked St.
  Leo I and St. Basil the Great both landing on/near a real octave in
  some year, and both are fully transferred, no commemoration at all —
  so the plain-suffix categories structurally can never be observed
  inside a real octave, only in the ordinary weeks around it.
- One practical check this settled without needing a code change:
  `common-of-pope-martyrs` (Ss. Soter & Caius, Ss. Cletus & Marcellinus)
  is itself one of the dedicated-chain categories, so its one stored
  file is technically the Paschaltide-only text, not a universal
  default — but tracing both saints' fixed dates against every possible
  Easter date confirms this is harmless: whenever their date would fall
  outside Eastertide, it lands inside the Sacred Triduum instead
  (always fully transferred, never rendered), so the stored text is
  correct for every real occurrence.

Built: `hours/resolve-common.ts`'s `appendPaschaltideAlleluia`/
`withPaschaltideAlleluia`, applied to the Common-category fallback
branch of `getBenedictusAntiphon` and `resolveMinorHourAntiphon` (not
the `-paschaltide`-file branch, which already had its own correct
wholesale text; not a saint's own unique `${propers}-antiphon`, which
would need its own live-requery to confirm one way or the other rather
than assuming the Common-category finding transfers; not capitula,
whose closing versicles get the same suffix in the reference engine too
— confirmed, St. Robert's Terce capitulum — but whose text is one
embedded R./V. blob per language rather than a clean appendable tail,
not attempted this pass). Guards against double-appending on a category
whose stored text already ends in "allelúja" (`common-of-pope-martyrs`,
per above). 4 new tests in `tests/hours/resolve-common.test.ts`
(`Paschaltide alleluia suffix on Common-category antiphons`), 300 tests
passing.

Verified end-to-end: a runtime smoke test (folded into
`tests/hours/resolve-common.test.ts`) covering the same-person reuse
case (Finding of St. Stephen → St. Stephen Protomartyr's own files),
Gregory's unique content, and a shared-Doctor case, plus 8 more spot
checks across the newly-wired families during the pass itself.

### The last 2 duplex-majus+ P/T/S/N gaps, and a final-sweep bug (2026-08)

Closes out `minorHoursCommon`/P/T/S/N entirely: the original 29-saint
pass held back 5 saints for lacking a collect at the time (`st-benedict`,
`all-saints-of-the-benedictine-order`, `st-michael`, `st-maurus-abbot`,
`ss-placid-and-companions`) — the latter two already got Common
fallbacks wired in earlier sections above, and St. Michael redirects to
Apparition of St. Michael's own content (see below). That left **St.
Benedict and All Saints of the Benedictine Order** as the only two
duplex-majus+ saints on the whole calendar with no P/T/S/N at all.
Live-queried both the same way the original 29 were: real, unique
proper text for each (Benedict's own Prime antiphon is the opening line
of St. Gregory the Great's *Life of St. Benedict*; All Saints of the
Benedictine Order draws on Matt 13:43, Luke 6:22-23, and a well-known
All Saints antiphon) — 14 new files total, English translated
throughout (`status.en: draft`) since the source rendered none for
either saint's own texts.

**A final calendar-wide sweep** (every saint's `resolveMinorHourAntiphon`/
`resolveMinorHourChapter`, synthetic-winner methodology) turned up one
more real gap: **St. Michael himself**. His `propers` field points at
Apparition of St. Michael's own id (the two feasts are word-for-word
identical, already established), and `getDayCollect`/
`getBenedictusAntiphon` both explicitly follow `propers` when looking
up their own content — but `getMinorHourOverrideId` doesn't; it keys
off the day's own winner id directly (`"st-michael"`, not
`"apparition-of-st-michael"`), so `resolveMinorHourAntiphon` was looking
for a `prime-antiphon-st-michael.yml` that was never going to exist.
Fixed the same way as Finding of St. Stephen: `minorHoursCommon:
apparition-of-st-michael` on his own record, reusing his delegate's
already-authored proper files directly. A real, narrow gap in the
mechanism itself (not just missing data) — worth remembering if any
other saint's `propers` ever points somewhere other than its own id.

**Every real saint on this app's calendar now has real P/T/S/N
content.** The only remaining absence is Ss. Ursula and Companions
(deliberately left unwired, no way to verify a category of one) and
the `example-confessor` test placeholder, which was never meant to
have any.

### Christmas-Octave Sunday content — done (2026-08)

Was: only the Benedictus antiphon existed for `christmas-octave-sunday`
(`christmas-octave-sunday-benedictus-antiphon.yml`) — no collect (a
separate, older file, `christmas-octave-sunday-collect.yml`, did already
exist and was untouched here), no Lauds psalmody override or office
bundle, no Prime/Terce/Sext/None content at all.

Real code gap found and fixed along the way: `getMinorHourOverrideId`
(`hours/resolve-common.ts`) only ever checked the *sanctoral* branch of
`resolveOfficeWinner`'s result — a named temporal feast in
`ALWAYS_OVERRIDE_TEMPORAL_IDS` (`marian-saturday`, `christ-the-king`) got
a Lauds psalmody override via `getPsalmodyOverrideFor`'s own separate
check of that set, but silently *never* a Prime/Terce/Sext/None one,
however much content got authored for it — nothing had actually
exercised that path until now. Fixed by mirroring `getPsalmodyOverrideFor`
`hours/lauds.ts`'s own temporal-id branch.

Content itself needed a different source than every other pull this
session: Monastic Tridentinum 1617 has no real "Dominica Infra Octavam
Nativitatis" identity of its own under its own calendar rules on most
Dec 26-29 Sundays (verified live: 2026-12-27 renders as "S. Joannis
Apostoli," not the Sunday, under `version=Monastic Tridentinum 1617` —
only the collect happens to be old enough to have been pulled from a
date, 2029-12-30, where 1617 *did* render it outright). Lauds psalmody,
office bundle, and Prime/Terce/Sext/None content instead pulled live
under `version=Divino Afflatu 1954` (2026-12-27), consistent with this
app's own stated precedent of following D.A. for the calendar/precedence
layer even while keeping Monastic 1617's shape for the hours themselves —
this is one of the rare days where that split genuinely can't be
maintained, since 1617 has nothing of its own to shape. `christmas-
octave-sunday-collect.yml`'s existing 1617-sourced text was kept as-is
(cross-checked identical against both the new D.A. query and the raw
Monastic Tridentinum 1617 source file, `Tempora/Nat1-0.txt`'s own
`[Oratio]`) rather than replaced.

Files: `lauds-psalmody-overrides/christmas-octave-sunday.yml` (3 psalms +
Canticle of the Three Young Men + Laudate, all distinct antiphons),
`lauds-capitulum-/-versicle-/-hymn-christmas-octave-sunday.yml` ("A solis
ortus cardine"). No `lauds-responsory-christmas-octave-sunday.yml`
authored — genuinely absent from the live rendering (hymn leads straight
into the versicle, no short responsory in between), not an oversight;
left "missing" honestly rather than invented. Prime/Terce/Sext/None all
four get a proper antiphon (a short "Allelúia, ..." set drawn from the
day psalter, distinct per hour); Terce/Sext/None each get a proper
chapter (Gal 4:1-2/4:4-5/4:7, matching the raw 1617 source's own
`Nat1-0.txt` exactly) with its own versicle per the D.A. live rendering
(not the 1617 source's own versicle, which reuses one versicle across
all three hours — a cross-era mismatch flagged here, not silently
papered over, since this whole day's content already mixes eras out of
necessity).

Not attempted: `Nat2-0` (Feast of the Holy Name of Jesus) isn't in this
app's sanctoral/temporal calendar at all yet — a separate, bigger ask if
wanted (see the original scoping discussion this session).

### Overlapping octaves — done (2026-08)

Was: `resolveOfficeWinner`/`getDayLabel` picked whichever active octave
started *first* whenever more than one was active — an accident of
insertion order, not a real precedence rule. St. Lawrence's octave (Aug
10-17) and the Assumption's (Aug 15-22) genuinely overlap on Aug 16-17
every year, so this wasn't hypothetical.

Design (per direct instruction, not independently re-derived from a
primary source the way most of this project is): when more than one
octave is active on a day with no temporal standing of its own to yield
to first, compare them by rank — highest wins outright, exactly like a
single octave already out-ranks a weak rival saint; every other active
octave still gets commemorated, never silently dropped. A rank tie goes
to whichever octave started more *recently* — reasoning given: day 1 of
a newly started octave needs to be fully present, which is the whole
point of an octave starting, so the older one yields instead. No real
tied-rank case exists yet in this app's calendar to verify that
empirically, unlike almost everything else in this project — flagged as
a stated, unverified-by-evidence design decision, not a live finding.

Also needed `OctaveConfig.closingDayRank` (default `duplex`): an
octave's own *closing* day ("in Octava," the 8th/final day) is elevated
above its ordinary in-between-days rank — live-verified as a real,
general pattern (not a St. Lawrence quirk): both his own Aug 17 and the
Assumption's own Aug 22 show as Duplex, above their otherwise-Semiduplex
ordinary days. This is *why* Lawrence's octave beats the Assumption's on
their one real overlap day (Aug 17) despite the Assumption being the far
higher-ranked feast overall (Duplex I. classis vs. Lawrence's Duplex II.
classis) — his closing-day elevation (Duplex) beats her ordinary day-3
strength (Semiduplex). Feeds both `strictestThreshold` (a rival saint
now needs Duplex+, not just Semiduplex+, to beat an octave specifically
on its own closing day) and the new octave-vs-octave comparison
(`calendar/octaves.ts`'s `resolveActiveOctave`/`pickWinningOctave`) —
same underlying "how strong is this octave today" question, two
different comparison partners.

`getDayLabel` was carrying its own separate, never-actually-correct
octave-selection logic that additionally never checked
`temporalCategory` at all — found while testing the real Aug 16 overlap
(a Sunday that year): it was showing an octave name even though the
*temporal* Sunday has real standing and wins outright, live-verified
counterexample being the Christmas Octave's own stack (Dec 30), which
was wrongly showing "3rd Day within the Octave of The Holy Innocents"
instead of the correct plain temporal label. Fixed by having
`getDayLabel` share the exact same `ordinary-feria` gate and
`resolveActiveOctave` call as `resolveOfficeWinner`, so the two can never
disagree about which identity governs a day.

### Dec 26-30's own Sunday rule — done (2026-08), a deliberate remix

Per direct instruction — a real, admitted departure from Monastic 1617
itself, not something independently verified live the way almost
everything else here is: within Dec 26-30 inclusive (every year contains
exactly one real Sunday somewhere in that 5-day window), whichever date
is the actual Sunday always celebrates Nat1-0 (the Sunday within the
Octave of the Nativity) outright, with **no rank check at all** — even a
Duplex-II-classis saint (Stephen/John/Innocents) is fully displaced, not
just commemorated. The displaced saint isn't commemorated in place
either; he reappears in full as the day's own winner on Dec 30 instead —
the block's one date with no fixed saint of its own. On a year Dec 30
itself is the actual Sunday, nothing is displaced and it resolves clean.

This intentionally diverges from what Monastic 1617 itself does on Dec
26-29 (there, live-verified, the Duplex-II-classis saint wins and the
Sunday is merely commemorated — see the December-sanctoral tests) and
from the general `privileged-sunday` transfer rule elsewhere in this app
(which only transfers ranks below duplex-majus and commemorates
everything at or above it in place) — this window transfers *every*
rank, full stop, and never commemorates the loser in place.

Implemented as its own dedicated, narrowly-scoped layer
(`calendar/index.ts`'s `applyChristmasOctaveSunday`, same architectural
pattern as `applyMarianSaturday`/`applyChristTheKing`) rather than by
changing the shared `decideOccurrence`/transfer machinery other Sundays
rely on — deliberately contained blast radius. Dec 31 and everything
outside Dec 26-30 is untouched; christmastide's Sunday there still uses
the plain `ordinary-sunday` category as before (St. Silvester still
beats the Sunday on Dec 31, live-verified, unaffected by this).

### Benedictus antiphons for temporal (non-sanctoral) days — done

`<id>-benedictus-antiphon.yml` in `data/propers/temporal/` — all 54
temporal ids now have one (`marian-saturday`/`christ-the-king` from
earlier, the 51 numbered Sundays/seasons done in one pass, plus
`christmas-octave-sunday` itself, corrected below).

**`christmas-octave-sunday` correction**: originally left unauthored
here, on the theory that every date this id can resolve to collides with
a specific octave-day saint. Wrong — the reference engine has its own
dedicated proper for this Sunday specifically, `Tempora/Nat1-0`
(`web/www/horas/Latin/Tempora/Nat1-0.txt` in the reference repo, and its
Monastic counterpart under `TemporaM/`), distinct from the individual
per-weekday octave pages (`Nat02`..`Nat09` = the ferias, keyed by day
offset from Christmas). Most years the Sunday does lose outright to a
Duplex-II-classis octave saint (St. Stephen/St. John/Holy Innocents,
whichever Dec 26–28 lands on) and only gets a bare commemoration — that's
what the earlier searches kept finding — but 2029-12-30 is a clean year
where the Sunday itself wins outright (`Dominica Infra Octavam
Nativitatis ~ Semiduplex Dominica minor`, no commemoration line at all),
so that's the date the content was pulled from. Benedictus antiphon:
"Dum médium siléntium" / "While all things were in quiet silence" — note
the source's own English translation genuinely omits the closing
"alleluia" that the Latin has; transcribed as-is, not smoothed over.

Method: rather than hunting for "clean years" one at a time, wrote a
script that iterates every day of a test year through this app's own
`resolveDay`, buckets each Sunday by which temporal id actually won, and
picks one real date per id straight from the app's own logic — then
cross-checked each date's *live* D.O. title for a saint collision before
trusting the content (a handful, mostly on privileged Eastertide/
post-Pentecost Sundays, needed a different year: this app's own
privileged-Sunday rule never lets a saint displace those Sundays, but the
live reference engine's calendar is bigger than this app's and does let
some by-date saint win there, e.g. St. Mark on Easter V some years, "Our
Lady of the Snows" — not modeled in this app at all — landing on
Post-Pentecost XI whenever the date lines up with Aug 5).

### Octave content — done (2026-08)

Was: `getDayCollects` skipped a `kind: 'octave'` commemoration entirely,
and every other office part (psalmody, hymn, chapter, Benedictus
antiphon) stayed keyed off the plain temporal day underneath the octave
— so an ordinary day within e.g. St. Lawrence's own octave showed nothing
of his at all, not even a label beyond the day-label line. Root cause:
found live (Tridentine 1910, 2026-08-12) that on a day like this, the
*whole* office — chapter/responsory/hymn/versicle, psalm antiphons,
Benedictus antiphon, day collect — comes from the octave's own feast, not
the plain temporal day; only a *weaker* commemorated saint on top of that
(St. Clare, that same date) gets a separate Ant+V/R+collect block of her
own. Fixed via `hours/resolve-common.ts`'s `resolveOfficeWinner`, used
everywhere `getDayCollect`/`getDayCollects`/`getBenedictusAntiphon`/
`getPsalmodyOverrideFor` used to key off `day.winner` directly — and
`getDayCollects` now renders a real Ant+V/R+collect bundle for a
sanctoral commemoration (`${propers}-commemoration.yml`, e.g.
st-clare-commemoration.yml) when authored, falling back to a *labeled*
bare collect (or a labeled "missing" placeholder) rather than the old
unlabeled one that just read "PRAYER … (translation pending)" with no
indication whose it was.

Remaining gap, now closed (2026-08): `resolveOfficeWinner` deliberately
only kicks in when `temporalCategory === 'ordinary-feria'` (the octave
day has no standing of its own) — the live-verified counterexample was
the Christmas Octave's own stack (Christmas + Stephen + John + Holy
Innocents, e.g. on Dec 30, `privileged-feria-minor`): the *temporal* day
itself keeps the office there ("Dominica Infra Octavam Nativitatis", `{ex
Proprio de Tempore}`), and each of the four octaves gets its own separate
"Commemoratio Octavæ …" Ant+V/R+collect block instead. Fixed by adding a
`kind === 'octave'` branch to `getDayCollects` (gated on
`day.temporalCategory !== 'ordinary-feria'`, the same condition
`resolveOfficeWinner` uses, so an ordinary octave day's already-rendered
primary content never gets a redundant repeat), plus one new
`${id}-octave-commemoration.yml` proper file per octave for the four
Christmas-stack ids (`christmas-day`, `st-stephen-protomartyr`,
`st-john-apostle`, `holy-innocents`), sourced from each Sancti file's own
`[Octava 3]` section in `../divinum-officium-reference` (Second Vespers'
own variant — the reference engine has three slightly different
Octava 1/2/3 texts per octave, roughly Vespers I/Lauds/Vespers II, but
this app's `getDayCollects` renders one shared block regardless of hour,
matching the pre-existing "one Orémus per commemoration, not chained"
simplification already noted above). The Holy Innocents' own English
Octava 3 block turned out to carry a mismatched antiphon in the reference
engine itself (the Common-of-Many-Martyrs "Isti sunt" text, not a
translation of the Latin antiphon printed beside it) — worked around by
sourcing the correct English from that same file's own `[Ant 3]`/
`[Versum 3]` sections instead (same Latin text, correctly paired
translation) — see that file's own comment. Live-verified end-to-end via
`tests/hours/vespers.test.ts`'s `2033-12-30` stacking case (all four
blocks resolve `verified` in both languages, not `missing`). This closes
the octave-content gap listed at the top of this file's "Open work"
section — no octave content work remains open.

## Benedictus antiphons for every sanctoral/temporal entry (in progress)

Full scope, corrected after actually enumerating it (an earlier pass here
under-counted): **109 sanctoral entries have `propers` set** (a saint id
that both a collect and an antiphon key off) and need a real Benedictus
antiphon; **94 more saints have `propers: null`** (no collect *or* antiphon
authored, ever) and are a separate, bigger prerequisite — see below. Plus
**~52 temporal entries** (every numbered Sunday/feria) need one too.

### Confirmed systemic bug, not just "unverified"

Every pre-existing `<saint>-antiphon.yml` file (there were 17, pulled during
the original sanctoral-content sessions, months before Lauds existed) was
sourced from the **Roman/secular rite's raw `[Ant 1]`**, not a live Monastic
1617 query. St. Lawrence's turned out wrong this way; going back and
correcting 3 more (Chair of St. Peter at Rome, St. Agnes, Conversion of St.
Paul) while working January confirms it's the same mistake every time, not
a one-off — **treat all pre-existing antiphon files as wrong until
re-verified against a live Monastic query**, not just flagged for a look.

### Method, worked out the hard way this pass

A saint's raw per-date file (`Latin/SanctiM/MM-DD.txt`, falling back to
`Latin/Sancti/MM-DD.txt`) is *not* a reliable shortcut — Monastic doesn't
always compute the Benedictus antiphon from a literal `[Ant 1]` field (major
feasts like Lawrence draw it from elsewhere entirely), and a shared date
with two co-assigned saints only has *one* real file, which may belong to
whichever saint reliably loses the day, not the one this app's calendar
means to ask about. What actually works: a live `command=prayLaudes` query
on the saint's own date, checked by eye against 2-3 different years for a
title that plausibly names the saint and stays consistent — full automation
kept producing false positives/negatives on exactly these edge cases, so
each result needs a human read, not just a script.

Also found along the way, now fixed: **this app's calendar and the
reference engine's own calendar don't always agree on dates.** St. Hilary
of Poitiers was assigned Jan 23 here; Divinum Officium's own calendar has
him fixed on Jan 14, confirmed directly against the raw `[Officium]` file
(Jan 23's own file is St. Raymond of Peñafort, not Hilary at all). The
original Jan-23 sourcing wasn't fabricated — 2029 is a real year where
Jan 14 lands on a Sunday and every day through the 22nd is already taken,
so Hilary genuinely transfers all the way to the 23rd *that one year* —
but the app had mistaken a transfer-year artifact for his real date. Moved
to Jan 14 in `sanctoral-calendar.yml`; his collect text (unaffected by
which day it's said on) re-confirmed identical; antiphon pulled fresh.
Worth a pass at some point checking whether any other already-authored
saint has the same kind of mismatch.

### Progress

- **January (9 total): done, all 9.** St. Marcellus I, Chair of St. Peter at
  Rome, St. Paul the Hermit, Ss. Fabian and Sebastian, St. Agnes, Ss.
  Vincent and Anastasius, Conversion of St. Paul, St. John Chrysostom, and
  St. Hilary of Poitiers (moved to his real date, Jan 14, per direct
  correction — see above).
- **February (7 total): done, all 7.** St. Agatha, St. Dorothea, St.
  Apollonia, St. Scholastica (translated, source had no English),
  St. Valentine, Chair of St. Peter at Antioch, St. Matthias (source's own
  English existed but was jarringly modern next to everything else's
  archaic register — reworded to match, marked `draft`).
- **March (6 total): done, 5 authored/corrected, 1 flagged.** St. Lucius I,
  St. Thomas Aquinas, the Forty Holy Martyrs, St. Gregory the Great
  (translated, source had no English — the "non Angli sed Angeli" antiphon),
  St. Joseph (corrected). St. Frances of Rome left `missing`: Simplex rank,
  and March 9 always falls within Lent every tested year (2026-2035) — she
  structurally can never win her own day outright under this app's own
  calendar rules, so there's no live-rendered antiphon to source at all.
- **April (8 total): done, all 8.** St. Francis of Paola, St. Leo I,
  Ss. Tiburtius/Valerian/Maximus, St. George, St. Mark, St. Vitalis,
  St. Robert, St. Peter Martyr.
- **May (13 total): done, all 13.** 3 corrected (Ss. Philip and James,
  St. John Before the Latin Gate, the Apparition of St. Michael), 10
  authored fresh (Athanasius, Monica, Stanislaus, Nereus/Achilleus/
  Domitilla/Pancras, Boniface, Ubald, Peter Celestine, Urban I,
  Eleutherius, John I). Urban I/Eleutherius/John I needed 2038 to find a
  year not governed by Ascensiontide/Pentecost/Trinity/Corpus Christi.
- **June (16 total): done, all 16.** 5 corrected (Ss. Basilides/Cyrinus/
  Nabor/Nazarius, the Nativity of St. John the Baptist, Ss. John and Paul,
  Ss. Peter and Paul, Commemoration of St. Paul), 11 authored fresh
  (Ss. Marcellinus/Peter/Erasmus, Ss. Primus and Felicianus, Barnabas
  [reworded English, same register fix as Matthias], Anthony of Padua,
  Basil the Great, Ss. Vitus/Modestus/Crescentia, Ss. Mark and Marcellian,
  Ss. Gervase and Protase, Paulinus of Nola, William Abbot, Leo II).
- **July (9 total): done, all 9.** 1 corrected (Mary Magdalene), 8 authored
  fresh (the Seven Brothers/Rufina/Secunda, Symphorosa and Her Seven Sons,
  Margaret of Antioch, Apollinaris, James the Greater [reworded English,
  same fix as Matthias/Barnabas], Anne, Nazarius/Celsus/Victor I/
  Innocent I, Abdon and Sennen).
- **August (12 total): done, all 12.** 1 corrected (Peter's Chains), 9
  authored fresh (Dominic, Ss. Cyriacus/Largus/Smaragdus, Assumption,
  Bernard, Bartholomew [reworded English, same fix as Matthias/Barnabas/
  James the Greater], Louis, Zephyrinus, Augustine), 2 already correct as
  found (Finding of St. Stephen, the Decollation of St. John the Baptist
  — both plain unmodified matches against the live query, no fix needed).
- **September (5 total): done, all 5.** 1 corrected (Nativity of the BVM),
  4 authored fresh (Matthew [reworded English, same fix as Matthias/
  Barnabas/James the Greater/Bartholomew], Ss. Cyprian and Justina,
  Ss. Cosmas and Damian, Jerome).
- **October (7 total): done, all 7.** Francis of Assisi, Callistus I,
  Luke [reworded English, same fix as Matthias/Barnabas/James the
  Greater/Bartholomew/Matthew], Ss. Chrysanthus and Daria, Vigil of
  Ss. Simon and Jude, Ss. Simon and Jude [same reworded-English fix],
  Vigil of All Saints.
- **November (7 total): done, all 7.** 2 corrected (Cecilia, Andrew), 4
  authored fresh (All Saints, Clement I, Catherine of Alexandria,
  Silvester Abbot). Martin of Tours already matched the live query as
  found, no change needed.
- **December (10 total): done, 9 authored/corrected, 1 flagged.** 2
  corrected (Lucy), 6 authored fresh (Nicholas, Immaculate Conception
  [minor English fix -- the source's own translation said "her Birth"
  for a Conception antiphon, an obvious copy-paste leftover from the
  Nativity-of-the-BVM formula this reuses; corrected, marked draft],
  Damasus, Thomas of Canterbury). Thomas Apostle, Stephen Protomartyr,
  John Apostle, and Holy Innocents already matched the live query as
  found, no change needed. St. Bibiana left `missing`: Simplex rank, and
  Dec 2 falls within Advent in every tested year 2026-2029 -- same
  structural block as St. Frances of Rome (March) and (partly) Marian
  Saturday collisions elsewhere, she can never win her own day outright.

**All 109 sanctoral entries with a `propers` id: done.** 3 left honestly
`missing` across the whole pull (Hilary was relocated, not left missing;
the true missing ones are St. Frances of Rome, St. Bibiana, and Ss.
Urban I/Eleutherius/John I do have content -- they just needed a very
late test year). Every pre-existing antiphon file that predated this
pass has now been checked against a live Monastic query; most needed
correction (confirmed Roman-rite sourced), a handful already matched.
- **All ~52 temporal entries: done, including `christmas-octave-sunday`**
  (see the section above for method and its own correction).
- **94 `propers: null` saints: done (2026-08)** — see the dedicated
  section below for the collect-authoring pass and the era-specific-
  calendar discovery it took to do reliably. Benedictus antiphon: real
  content for all of them except the deliberate `example-confessor` test
  placeholder and Ss. Ursula and Companions (left honestly unset, no way
  to verify a category of one — see her own file's comment) — 66 via a
  shared Common (own dedicated section below); of the other 26 (own
  proper collect), 21 share an existing Common (7 live-query-verified
  directly, 14 more resolved straight from source once "unwinnable"
  stopped being treated as a content blocker — see that section's own
  note) and 5 have real unique text (one, Ss. Euphemia/Lucy/Geminian,
  honestly marked `draft` rather than `verified` since it can never be
  checked against a live rendering). P/T/S/N: done for all 94, and for
  the whole rest of the calendar besides — see "`minorHoursCommon` gap
  swept across the whole calendar" and "The last 2 duplex-majus+ P/T/S/N
  gaps" below; every real saint on this app's calendar has real P/T/S/N
  content now except Ss. Ursula and Companions, same reason as above.

### Bug found and fixed while pulling christmas-octave-sunday's antiphon

`data/calendar/temporal-categories.yml` had christmastide's own Sunday
coded `privileged-sunday` (Advent's tier — never displaced short of
Duplex-majus+, and even then only commemorated in return). Live-checking
all three Duplex-II-classis "Comites Christi" collision dates (St.
Stephen/John/Holy Innocents falling on the Sunday, plus St. Silvester,
plain Duplex) showed the exact reverse: the saint wins outright every
time, the Sunday only commemorated — `ordinary-sunday`'s rule, not
`privileged-sunday`'s. Fixed by reclassifying christmastide's `sunday` to
`ordinary-sunday` (the `feria` side, `privileged-feria-minor`, was
already correct and untouched). New tests in
`tests/calendar/december-sanctoral.test.ts` cover both directions: the
saint winning on the three verified collision dates, and St. Thomas
Becket (Semiduplex) still transferring off the Sunday rather than
winning, unchanged by the fix.

### Bug found while checking February's results — fixed (2026-08)

`hours/resolve-common.ts`'s `splitNamedAntiphon` always marked its output
`verified` regardless of the input status (`verifiedText()` hardcoded it) —
so a `draft`-marked antiphon (Scholastica's translation, Matthias's
reworded English, and every other reworded/translated antiphon from the
sanctoral pull) rendered as if fully verified everywhere it's actually
displayed — silently dropping the "unverified draft text" UI marker
(`src/ui/styles.css`'s `.text-draft`, a dashed underline + tooltip) — even
though the source YAML correctly said `draft`. Affected every split
antiphon across every hour (Prime, Terce, Sext, None, Compline, Lauds),
not just Scholastica/Matthias — pre-existing, just never surfaced until
content with a real `draft` status went through it.

Fixed by having `splitNamedAntiphon` take a real `ResolvedText` (not just
bare text) and carry its actual per-language status through to both the
incipit and full outputs. Callers that only ever had a plain
`Partial<Record<string,string>>` with no status of its own to begin with
(the plain weekday-default antiphons, Lauds psalmody overrides — neither
type tracks a status field, always implicitly verified by their own
live-checked file convention) now wrap with `verifiedText()` explicitly
at the call site instead of that being silently baked into
`splitNamedAntiphon` itself. `resolveMinorHourAntiphon` (Prime/Terce/
Sext/None's own override lookup) now returns a full `ResolvedText` too,
for the same reason. Verified end-to-end with St. Scholastica's own
draft-English antiphon (`tests/hours/resolve-common.test.ts`).

## Mechanism not built at all

- **Vespers ordo — built (2026-08)**: `src/hours/vespers.ts` now resolves a
  full ordo (opening versicle, weekday/seasonal psalmody with the 115+116
  join and 138 split, chapter/responsory/hymn/versicle bundle with
  Advent/Lent/Passiontide/Paschaltide overrides, Magnificat with weekday-
  default antiphon fallback, day-collects) off `resolveEveningDay`, so
  First/Second Vespers precedence and the cross-day commemoration
  (`commemorationOf` — "the loser is commemorated unless the loser is
  simplex", live-verified) both apply before any part runs. See
  `data/hours/vespers.yml`'s header and `tests/hours/vespers.test.ts`.

- **Vespers cross-day commemoration now labeled, and feast-side privilege
  now has a mechanism (2026-08-26)**: a cross-day commemoration
  (`Commemoration`'s `vespersNote: 'today' | 'tomorrow'`, set only by
  `calendar/vespers.ts`'s two `commemorationOfDisplaced*` functions) now
  renders as "(of today)"/"(of tomorrow)" in English, "de præcedenti"/"de
  sequenti" in Latin (`calendar/day-label.ts`) — the real Divino Afflatu
  phrasing, per `horascommon.pl`. Also added `forbidsSuccessorCommemoration`
  to both `SaintRecord` (`calendar/feasts.ts`) and `TemporalFeastRecord`
  (`calendar/temporal-feasts.ts`) — the feast-side mirror of the day-side
  privilege `calendar/commemorations.ts` already modeled (a privileged
  Sunday/major-feria can refuse a weak commemoration *back*; nothing
  previously let a high-ranked feast refuse to commemorate *forward* at
  its own kept Second Vespers). Deliberately a flag, not a ported
  rank-threshold formula (the reference engine's own version is a messy
  fractional-rank matrix) — set only after checking a specific date
  against the live engine, same convention as `vigilOf`.
  - The concrete case that motivated this: **Most Precious Blood (July 1,
    Duplex I classis) vs. the Feast of the Sacred Heart** (Easter + 68
    days), which only ever fall adjacent in a very-late-Easter year
    (Easter in late April). Read the Perl variables carefully (initial
    pass here got this backwards): it's **Sacred Heart's own kept Second
    Vespers that refuses to commemorate Precious Blood** the next
    evening ("Vespera de præcedenti; nihil de sequenti", `$winner` =
    Sacred Heart/`Pent02-5` = *today* at that point in the branch,
    `$cwinner` = Precious Blood/`07-01` = *tomorrow*), not the reverse.
    The other direction — Precious Blood's own Vespers commemorating a
    Sacred Heart that happens to fall the day *after* it, the other rare
    late-Easter arrangement — has no matching carve-out in the reference
    engine at all, so `forbidsSuccessorCommemoration` is deliberately set
    only on `sacred-heart.yml`, not `most-precious-blood.yml`.
  - Still open, separately: the reverse gap remains in daytime
    `decideOccurrence` (`calendar/commemorations.ts`) — a sanctoral
    winner there unconditionally gets its displaced Sunday/feria
    commemorated back, regardless of the winner's own rank. Not addressed
    here; this pass only covered the Vespers cross-day direction.

- **Ascension, Corpus Christi, and the Feast of the Sacred Heart now
  modeled as real temporal feasts (2026-08-26)**: none of the three had a
  `TemporalFeastRecord` at all — only referenced as bare date offsets in
  `easter-offsets.yml`/`temporal-categories.yml`/
  `isMajorFixedFeastOfTheLord`. Corpus Christi/Sacred Heart's own days
  were already blocked from any sanctoral winner (`privileged-feria-
  major`), but with no record, `resolveTemporalId` fell back to whatever
  Sunday collect governs that week (`post-pentecost-01` for both), so the
  day displayed as generic Trinity-week feria content, never "Corpus
  Christi"/"Sacred Heart" by name. Ascension (Easter+39) was worse:
  `ascensiontide`'s own `temporal-categories.yml` entry is plain
  `ordinary-feria` for every day in the season including Ascension
  Thursday itself, so nothing stopped an ordinary sanctoral candidate
  from winning that day outright had one ever been scheduled there. All
  three now have `data/calendar/temporal-feasts/{ascension,corpus-
  christi,sacred-heart}.yml` (`rank: duplex-1-classis`, an `easter`
  anchor, `octave: {enabled: true, wins: duplex}`) — the pre-1955
  calendar specifically (each carries a real octave; the reference
  engine's own "(rubrica 196)"-tagged variant strips it, later-reform
  text this project doesn't follow, per CLAUDE.md). Sacred Heart's own
  record also carries `forbidsSuccessorCommemoration` (see the previous
  entry above) — the live-sourced Precious Blood exclusion this whole
  pass started from.
  - `octave.wins: duplex` is matched to `pentecost-sunday.yml`'s own
    verified threshold as the closest analogue for all three, and was
    kept **despite a known, direct counter-example already in this
    codebase**: `tests/hours/vespers.test.ts`'s St. Ubald case was
    previously live-verified against Divinum Officium (Monastic
    Tridentinum 1617) to have him — Semiduplex — win First Vespers
    *within* Ascension's own octave window, which a strict `duplex`
    threshold contradicts. Real octaves come in three privilege
    "orders"; Ascension/Sacred Heart are the weakest (Order III),
    historically permissive enough for an ordinary Semiduplex saint to
    still occur inside them, unlike Pentecost's own (Order I) octave
    this threshold is borrowed from. Kept at `duplex` anyway per direct
    instruction — the Ubald test (and St. Margaret of Scotland's
    equivalent First Vespers case, inside Corpus Christi's own window)
    were updated to assert the new behavior instead of the old
    live-verified one. Revisit if this turns out to misfire elsewhere;
    see `ascension.yml`'s own comment for the full note.
  - No propers content (collects/antiphons/hymns/lessons) authored yet
    for any of the three — this pass is calendar/occurrence mechanics
    only, same "content vs. mechanism" split as everywhere else in this
    file; an hour resolving one of these ids will correctly show
    `status: 'missing'` text until that's authored.

### Matins ordo — mechanism built, small content slice (2026-08)

`src/hours/matins.ts` was the last stub hour. Built as a genuinely different
shape from every other hour: no static `data/hours/matins.yml` parts list —
the ordo is assembled programmatically per day, since the real structure (1
nocturn on a plain ferial day, 3 on a Sunday or Duplex-and-higher feast,
with a *variable* number of readings) doesn't fit the static-array pattern
every other hour uses.

**Explicit, direct-instruction departures from the historical office** (see
memory `vu-not-a-reconstruction`/`vu-matins-design` — this app was never
meant to reconstruct any one historical form, Matins least of all):
- **Nocturn 1 is always the user's own continuous scripture-reading plan**
  (`src/propers/bible-plan.ts`, sourced from a personal TSV outside this
  repo), never the historical per-day lectionary — a variable number of
  readings per day, and deliberately *not* the Rule of Benedict's own
  "summer" reading contraction (Low Sunday–1st Sunday of November): this
  app reads in full year-round, live-confirmed the reference engine's own
  Monastic 1617 data does apply that contraction (`monastic.pl`'s
  `brevis_monastic`/`legend_monastic`) before deciding not to reproduce it.
- **Recombine, don't reproduce the historical lesson split**: where the
  source splits one continuous work across several numbered lessons purely
  to fill a fixed lesson-count slot, this app joins them into one reading
  (`src/propers/octave-readings.ts`'s `resolvePassages`, now reused outside
  the octave-day store it was built for — extended via the new
  `src/propers/nocturn-readings.ts`). Split only kept where the underlying
  source genuinely changes (e.g. a Gospel pericope vs. the homily on it).
- **Nocturn 3 gated at Duplex-and-higher**, plus every Sunday
  unconditionally — the user's own chosen threshold (real Monastic 1617 is
  more permissive), reusing the already-existing `isDoubleOrHigher` (which
  already means rank ≥ `duplex`, so no new rank helper was needed).
- **Generous multi-source reading inclusion, not winner-takes-all**: every
  commemorated saint (not just the office winner) and every active octave
  contributes its own Nocturn 2/3 reading when authored — mirrors
  `getDayCollects`'s existing "one collect per commemoration" pattern,
  applied to readings. Wired `src/propers/octave-readings.ts`'s
  `getOctaveReading` into a real caller for the first time — it had zero
  callers anywhere in the codebase before this, clearly built ahead of
  Matins and left unwired until now.
- **Gospel sourcing has exactly two allowed sources**: the user's own plan
  (flagged via `isGospel`, book-derived — confirmed the user's plan never
  assigns one on a Sunday, a deliberate editorial choice, not a gap) and
  the day's own genuine *proper* Gospel+homily. **A Common-of-Saints
  fallback Gospel is never used** — `nocturn-readings.ts` has no
  Common-fallback mechanism at all (unlike `collectCommon`/
  `benedictusCommon` elsewhere), so the exclusion falls out of the store's
  own shape rather than needing special-case code.
- **Invitatory (Ps 94) framed like an ordinary antiphoned psalm** — one
  opening (incipit-or-full), one full repeat after — not the historical
  interspersed-refrain shape (confirmed live: real practice repeats the
  antiphon between every verse group). Direct instruction, not an
  oversight.
- **Responsories matched loosely by scriptural book**, not exact citation —
  per the user's own description of how they actually worked historically.
  New seeded pool, `data/hours/matins-responsories-by-book.yml` +
  `src/propers/matins-responsories.ts` (currently just `isa`, from a live
  Advent-ferial query — a real seed, not a full pool).

**Sunday's fixed 12-psalm/3-canticle psalmody** (`data/hours/matins-sunday-
antiphons.yml`) — Nocturn 1 = Ps 20-21/22-23/24-25, Nocturn 2 = Ps
26-27/28-29/30-31 (Rule of Benedict ch. 18's own arrangement, distinct from
Roman Matins' 9-psalm Sunday scheme), Nocturn 3 = 3 OT canticles (Isa
33:2-10, Isa 33:13-18, Sir 36:14-19) under one shared antiphon. Transcribed
from the reference engine's own Monastic-labeled block (`Psalterium/Psalmi/
Psalmi matutinum.txt`'s `[Daym0]`), then live-verified end-to-end
(`command=prayMatutinum`, 2026-09-06) — psalm numbers, canticle citations,
and the 12-lesson/Te Deum sequence all matched exactly. Real gap found
along the way: the reference engine's own UI flags Monastic 1617 Matins as
"still incomplete and under construction" — the pieces actually queried
checked out regardless, but worth remembering if a future pull comes back
looking wrong.

**Real content authored for two clean proof dates** (one per branch, per
this project's usual "mechanism first, minimal live-verified proof, content
later" build order):
- **Ferial**: 2026-12-15 (Tuesday of Advent III) — the user's own plan's
  "Isa 36-37, Wis 14" (citation-only for now; full Vulgate/Douay-Rheims
  import is a deferred separate project, see below), matched against the
  seeded Isaiah responsory pool.
- **Sunday**: 2026-09-06 ("14th Sunday after Trinity", `post-pentecost-15`
  — the Trinity-counted display label and the Pentecost-offset storage id
  are a fixed, already-computed one-off relationship, not something new
  built here) — Nocturn 1 from the user's plan ("Tob 1-2, Sir 45",
  citation-only), Nocturn 2 a real, fully-translated patristic reading
  (St. Gregory the Great, *Moralia in Job*, Bk. 9 — live-transcribed, both
  languages), Nocturn 3 the day's own proper Gospel (Luke 7:11-16, the
  raising of the widow's son at Naim) plus St. Augustine's homily on it
  (Sermon 44), also fully transcribed. Real seam flagged, not papered over:
  this Sunday's *real* historical Nocturn 1 is the Book of Job (confirmed
  live) — Nocturn 2's Job commentary is kept as-is even though this app's
  own Nocturn 1 no longer reads Job that week, since it's real, valuable,
  season-proper patristic content, not tied to any specific saint.
- Two Isaiah chapters actually populated in `src/data/scripture/` (33,
  live-transcribed for the Sunday canticles) — small, real progress against
  the eventual bulk Bible import, not just placeholders.

Verified end-to-end: `tests/hours/matins.test.ts`, both branches, resolving
real dates through the actual `resolveOrdo` mechanism (psalm/canticle
counts, Te Deum presence, reading citations and statuses, Gospel flagging
and its Common-fallback exclusion, responsory matching) — not just checking
the YAML files exist. `npm test` (359 tests) and `npm run build` both pass.

### Reading-pool redesign — done (2026-08)

**Corrected a miscommunication in the original mechanism build above**: readings are now
pool-assembled, not fixed-slotted. Every source that can contribute for a day — the office
winner, every commemorated saint, a transferred-in feast (already folded into `day.winner`
by `resolveDay`, no separate lookup needed), every active octave, the plain temporal id, and
the user's own scripture-plan reading — is gathered into one ordered pool
(`buildReadingPool` in `src/hours/matins.ts`), then sliced into however many nocturns the
day's psalmody has (`distributeIntoNocturns`, ceiling division, pool order preserved). No
reading kind is pinned to a specific nocturn anymore — scripture is no longer reserved for
Nocturn 1, patristic content no longer reserved for Nocturns 2-3.

The 1-vs-3-nocturn psalmody gate (Duplex+/Sunday) is unchanged — it governs psalm/canticle/
Te Deum structure only. It no longer limits which days get patristic content authored or
shown: a sub-Duplex ferial day's single nocturn now pools in patristic/hagiographic content
whenever it's been sourced, alongside the scripture-plan reading — direct instruction, "we
want patristic readings for every saint where we can source one, not just duplex+."

`src/propers/bible-plan.ts`'s `getBiblePlanReadings` is now dual-keyed: most rows still key
off `(temporalId, weekday)`, but the source TSV's Dec 25 - Jan 13 stretch keys off a fixed
`MM-DD` calendar date instead (read straight through regardless of which temporal Sunday
governs the day) — a date can match both keys at once (e.g. a Sunday after Epiphany landing
inside Jan 1-13), and both are pooled rather than one overriding the other. No calendar-date
rows are authored yet (mechanism only); see memory `vu-matins-design` for the full design
writeup. `npm test` (359 tests, unchanged — the two proof dates still resolve the same
content, just via the new pool/distribute path) and `npm run build` both pass.

**Not done in this pass — explicitly deferred, not silently dropped:**
- Full Sunday/feast-calendar patristic-reading authoring — now framed as
  "every saint where a reading can be sourced," not just the ~75 Duplex+
  saints that get their own Nocturn 3 slot (see the reading-pool redesign
  above: content is no longer rank-gated, only nocturn *count* is). **Update
  (2026-08): the post-Pentecost season side of this is now done — see
  "Temporal-cycle nocturn-readings sweep" below.**
- The per-book responsory pool beyond its one seeded book (`isa`).
- The bulk Vulgate (Latin) + Douay-Rheims (English) import into
  `src/data/scripture/` — in scope for the project long-term, explicitly
  deferred past this pass per direct instruction.

### Temporal-cycle nocturn-readings sweep — post-Pentecost season done (2026-08)

The user noticed 2026-08-23's Matins (a plain Sunday, `post-pentecost-13`) was missing its
Nocturn 2/3 patristic reading entirely. Root cause: `post-pentecost-15` (above) was the *only*
temporal id with any `nocturn-readings` content authored — every other ordinary Sunday/week
across the whole year had none, a gap distinct from (and larger than) the sanctoral-saint gaps
tracked elsewhere in this file.

Scope turned out bounded: `calendar/temporal-id.ts`'s `resolveTemporalId` maps every day in a
liturgical week (Sunday and its ferias alike) to the same id, and `hours/matins.ts`'s
`buildReadingPool` already keys nocturn-readings lookup by that shared id — so **one file per
temporal id covers its entire week**, not one file per date. Authored all 23 remaining
`post-pentecost-NN` ids (01-14, 16-24; `-15` already existed) from the reference engine's own
`Tempora/PentNN-0.txt` (Latin) / `English/Tempora/PentNN-0.txt` files:

- **`post-pentecost-01` (Trinity Sunday, Duplex I classis)**: has its own full 9-lesson proper
  office, unlike every other Sunday in the season — Nocturn 2 (St. Fulgentius on the Trinity)
  plus *two* separate Nocturn 3 readings (St. Gregory Nazianzen's homily on Matt 28, and a
  distinct "Commemoratio Dominicæ" — St. Augustine's homily for the displaced First Sunday
  after Pentecost's own Gospel, Luke 6:36-42, which the Duplex I classis feast bumps into a
  commemoration rather than dropping — consistent with this project's generous-commemorations
  principle).
- **`post-pentecost-02` through `-11`**: each has real, unique proper content for *both*
  Nocturn 2 (a patristic/theological excerpt, one continuous author per week — St. John
  Chrysostom, St. Augustine, Pius XI, St. Ambrose, St. Jerome, St. Gregory the Great, etc.) and
  Nocturn 3 (that week's own Gospel pericope + homily). `-11` is the last week in the source
  with its own proper Nocturn 2; every week after it has Nocturn 3 only.
- **`post-pentecost-12` through `-14`, `-16` through `-24`**: Nocturn 3 only (no proper
  Nocturn 2 in the source from here on — the historical Nocturn 1/2 for these weeks is a
  continuous Old Testament book reading with commentary that isn't written out per-Sunday in
  the static files the way the Gospel+homily slot is; sourcing that properly would need
  live-CGI queries per date, out of scope for this pass, left as a known gap rather than
  guessed at). Every week's Lectio7 in the source is a bare "Et réliqua" Gospel incipit, not a
  full pericope — per this project's standing convention, not authored as its own Gospel
  reading; only the homily (Lectio7's own opening paragraph plus Lectio8-9, one continuous
  work) is authored, `isGospel: false`.
- **Responsory reuse discovered and followed**: the source itself cross-references the *same*
  two responsories for the Nocturn 2 and Nocturn 3 homily slots in nearly every week
  (`@Tempora/Pent01-2:ResponsoryN` for Nocturn 2, `@Tempora/Pent01-0:Responsory8` — "Duo
  Séraphim" — for the Nocturn 3 homily), rather than a fresh responsory each week. Resolved
  those cross-references once (from `Pent01-2`/`Pent01-3`/`Pent01-0`) and reused the resulting
  text verbatim across every file that points at them, rather than re-deriving per week.
- **A genuine authoring bug found and fixed mid-pass**: three of the first Nocturn-3-only
  files (`post-pentecost-13`, `-14`, `-16`) initially omitted the homily's own opening
  paragraph — it lives inside the source's `[Lectio7]` section (right after the bare Gospel
  incipit), not `[Lectio8]`, and an early extraction pass started from Lectio8 by mistake.
  Caught by re-diffing each file's opening sentence against the source before moving on;
  fixed by prepending the missing paragraph (Latin + English) to all three. Worth remembering
  for any future pass over this same source shape: always check that a Nocturn 3 homily
  reading's first sentence matches Lectio7's own prose, not Lectio8's.

Live-verified via `npx tsc --noEmit`, `npm test` (452 tests, up from 447 — 5 new cases in
`tests/hours/matins.test.ts`'s new "post-Pentecost season nocturn-readings sweep" describe
block, including a spot-check of 2026-08-23 itself), and `npm run build`. No fixture
collisions from this pass (pure content addition under already-existing ids, no calendar
changes).

**Not done in this pass — explicitly deferred, not silently dropped:**
- The rest of the temporal cycle: Advent (`advent-1..4`), Christmastide
  (`christmas-octave-sunday`), post-Epiphany (`post-epiphany-1..6`), Septuagesima through Lent
  (`septuagesima`, `sexagesima`, `quinquagesima`, `lent-1..4`), Passiontide/Holy Week
  (`passion-sunday`, `palm-sunday`), and Eastertide (`easter-sunday`, `easter-octave`,
  `easter-3..6`, `sunday-after-ascension`, `pentecost-sunday`) — same shape of gap, same
  pipeline, just not yet worked through season by season.
- Sourcing real Nocturn 2 content for `post-pentecost-12` onward via the Pentecost-offset id
  itself — **resolved differently, not left open**, see "Calendar-month/week nocturn-readings
  import" immediately below: that content turned out to be keyed by civil calendar month, not
  Pentecost offset, at the source, so it needed a second id scheme, not a live-CGI lookup.
- Ferial-specific (non-Sunday) content within a post-Pentecost week — not applicable here,
  since the whole week already shares one temporal id/file by design (see above); flagged only
  so a future reader doesn't go looking for `post-pentecost-13-tuesday.yml`-style files that
  were never meant to exist.

### Temporal-cycle nocturn-readings sweep — rest of the year done (2026-08)

Extended the post-Pentecost sweep above to the other 26 temporal ids — Advent (`advent-1..4`),
`christmas-octave-sunday`, post-Epiphany (`post-epiphany-1..6`), Septuagesima through Lent
(`septuagesima`, `sexagesima`, `quinquagesima`, `lent-1..4`), Passiontide/Holy Week
(`passion-sunday`, `palm-sunday`), and Eastertide (`easter-octave`, `easter-3..6`,
`sunday-after-ascension`) — using the reference engine's own per-season Tempora files
(`Adv1-0.txt`..`Adv4-0.txt`, `Nat1-0.txt`, `Epi1-0a.txt`/`Epi2-0.txt`..`Epi6-0.txt`,
`Quadp1-0.txt`..`Quadp3-0.txt`, `Quad1-0.txt`..`Quad6-0.txt`, `Pasc0-0.txt`..`Pasc7-0.txt`).
24 of 26 ids got real, sourced Nocturn 2/3 content on the first pass; the remaining 2 temporal
ids in this season range (`easter-sunday`, `pentecost-sunday`) were initially misdiagnosed as
genuine source gaps (see below) — corrected the same day, both now authored too.

- **`post-epiphany-1`'s real source is `Epi1-0a.txt`, not `Epi1-0.txt`**: the reference
  engine reassigns `Epi1-0.txt` to the later (1893/1921) feast of the Holy Family, which this
  project doesn't model (see the temporal-cycle survey note on Holy Family as a still-open
  pre-1955 gap, elsewhere in this file). `Epi1-0b.txt` (a thin `Dominica I post Epiphaniam`
  wrapper cross-referencing `Epi1-0a` for its own Nocturn 1) confirmed `Epi1-0a.txt` carries the
  real older "Dominica infra Octavam Epiphaniæ" office underneath. That file also has a real
  structural gap of its own: its `[Rule]` reads "Initia cum responsory" and it carries no
  `[Responsory4]`-`[Responsory9]` tags at all — its Nocturn 2/3 responsories come from a
  continuous weekly "Historia" cycle not transcribed per-Sunday in the static files (same class
  of gap as the post-Pentecost season's un-sourced Nocturn 2 weeks). `post-epiphany-1.yml`'s
  two readings are authored without a `responsory` field for this reason, not with an invented
  one.
- **`sunday-after-ascension`'s source file (`Pasc6-0.txt`) has no responsories of its own
  either** (only `[Responsory8]`, proper to a single lesson) — its own `[Rank]`/`[Rule]` say
  "ex Tempora/Pasc5-4" (Ascension Thursday itself), meaning the whole office, responsories
  included, is inherited from that day. Resolved by pulling the representative responsories
  directly from `Pasc5-4.txt` rather than leaving them unauthored.
- **`easter-sunday` (`Pasc0-0.txt`) and `pentecost-sunday` (`Pasc7-0.txt`) were initially
  (wrongly) called source-level gaps, then corrected same day.** Both files' own `[Rule]` really
  does read "1 nocturn" (their whole night office collapses to 3 lessons, straight to Te Deum) —
  true, and unlike every other Sunday in this sweep. But "1 nocturn instead of 3" is not the same
  claim as "no Nocturn 2/3 content exists": `Lectio1` in both files opens with a bare Gospel
  incipit ("Et réliqua") immediately followed by a full patristic homily (Easter: Pope St.
  Gregory the Great, Homily 21 on the Gospels, on Mark 16:1-7; Pentecost: the same, Homily 30, on
  John 14:23-31) continuing through Lectio2-3 — real, substantial, translatable content, just
  compressed into one nocturn rather than split across three the way an ordinary Sunday's 9
  lessons are. `hours/matins.ts` already supports single-nocturn source content tagged
  `nocturn: 1` (the Ember days' own nocturn-readings files do exactly this), and this app's own
  `threeNocturns` rule (every Sunday renders 3 nocturns regardless of the historical rubric for
  this particular feast) reslots the pool across 3 rendered nocturns anyway — the `nocturn: 1`
  tag only affects pool-ordering relative to other same-day contributors, not final placement.
  Authored `easter-sunday.yml`/`pentecost-sunday.yml` the same way as every other Nocturn-3-style
  entry in this sweep: incipit dropped, Lectio1's homily portion combined with Lectio2-3 into one
  reading, Responsory1 (the only nocturn-proper responsory transcribed in either source —
  Responsory3 doesn't exist, the office runs straight to the Te Deum after Responsory2's Gloria
  repeat).
- **Cross-reference reuse continues the post-Pentecost pattern**: several Sundays'
  `[Responsory4]`/`[Responsory7]` tags are themselves cross-references to a different file
  (e.g. `post-epiphany-3` through `-6` all point back to `post-epiphany-2`'s own Responsory4/7;
  `easter-6` points back to `easter-5`'s Pasc4-0 responsories) rather than fresh text — resolved
  and transcribed directly, same as before, not left as unresolved pointers.
- **The "homily begins inside Lectio7 itself" pattern, not just at Lectio8, held for nearly
  every Sunday in this pass** (not only the few post-Pentecost weeks that showed it) — every
  file's Nocturn 3 combines Lectio7's homily portion (dropping only the bare Gospel incipit)
  with Lectio8-9, using Lectio7's own responsory to represent the combined reading.

Live-verified via `npx tsc --noEmit`, `npm run build`, and `npm test` (806 tests, up from 777 —
a "rest-of-year temporal nocturn-readings sweep" describe block in `tests/hours/matins.test.ts`
covering all 28 ids end-to-end, `easter-sunday`/`pentecost-sunday` included; two pre-existing
tests in the ferial-branch describe block also needed a small filter update, since Advent I's
own ferial dates now legitimately pool in `advent-1.yml`'s patristic content alongside the
user's bible-plan readings). No fixture collisions (pure content addition under already-existing
ids, no calendar changes) except one real interaction the new content surfaced: `2026-02-23`,
initially picked as a `lent-1` proof date, turned out to actually be the Vigil of St. Matthias
transferred onto that Sunday and winning outright — swapped for `2031-03-03`, a clean `lent-1`
date.

**Temporal-cycle nocturn-readings sweep, all 52 ids, is now fully complete — no remaining gaps.**
The lesson from the easter-sunday/pentecost-sunday misdiagnosis: "the source's own `[Rule]` says
1 nocturn" is a claim about nocturn *count*, not about whether content exists — always check
whether `[Lectio1]`-`[Lectio3]` (or whichever lessons exist) carry real prose before calling a
short nocturn structure a content gap.

### Calendar-month/week nocturn-readings import — done (2026-08)

Follow-up to the post-Pentecost sweep above: the user asked about the "Nocturn 3 only, from
`post-pentecost-12` onward" gap it left open, correctly guessing that the reference engine's
real Nocturn 2 content for those later Sundays lives in older, month-labeled files rather than
being genuinely absent.

Confirmed: `Tempora/08N-0.txt` through `11N-0.txt` (N = 1-5, August-November, ~20 files) — the
old "Scriptura occurrens" reading cycle, "Dominica I-V Augusti/Septembris/Octobris/Novembris."
Each has `[Lectio1]`-`[Lectio3]` (Nocturn 1 scripture — **not imported**, the user's own
bible-plan store already owns that slot) and `[Lectio4]`-`[Lectio6]` (Nocturn 2 patristic
commentary — the actual target).

**The real complication, and why this needed a mechanism addition, not just content files**:
which month-week file governs a given Sunday is *not* a fixed `post-pentecost-NN`
correspondence — it's computed straight from the civil calendar (first Sunday of each month,
with irregularities), and shifts every year with Easter's date. Ported the reference engine's
own `Date.pm::monthday()` sub faithfully into `src/calendar/month-week-id.ts` (`monthWeekId`),
using this app's pre-1955/Tridentine rubric baseline (`Date.pm`'s own `$modernstyle = false`):
no 1960-rubric October week-III vanishing (skipped entirely, since that whole adjustment is
modernstyle-gated); November's backward-from-Advent renumbering for week > 0 *does* still
apply (that part of the source isn't modernstyle-gated). Live-verified against the localhost
CGI instance across 3 different years (2026, 2027, 2028) including the August 1st-falls-on-a-
Saturday skip case and November's real week-II-vanishes case (2026: Nov 1/8/15/22 render "I.
/III./IV./V. Novembris" — II is genuinely skipped that year, matching the ported algorithm
exactly).

Wired in as a **third, independent pooling source** in `hours/matins.ts`'s
`nocturnReadingIds`/`buildReadingPool` — `month-week-<id>` (e.g. `month-week-081`), pooled
alongside (not replacing) the ordinary `temporalId`-keyed lookup, same dual-key precedent as
`propers/bible-plan.ts`'s Dec25-Jan13 calendar-date stretch. All 20 ids authored as
`data/propers/nocturn-readings/month-week-<MMN>.yml`; each file's 3 lessons are one continuous
work (recombined per the standing rule), so one `NocturnReading` per file. `month-week-092`
turned out to cite the exact same Moralia-in-Job passage already authored for
`post-pentecost-15` — reused that existing translation rather than re-transcribing it.

Live-verified via `npx tsc --noEmit`, `npm test` (460 tests, up from 452 — a new "calendar-
month/week nocturn-readings import" describe block in `tests/hours/matins.test.ts` spanning
2026-2028 to prove the date-driven lookup, not a hardcoded id), and `npm run build`. No fixture
collisions (pure content + one new pooling source, no calendar changes).

**Not done in this pass**: the same month-week treatment for whatever Sundays this pass didn't
reach (there are none within Aug-Nov — all 20 files are authored); the rest of the temporal
cycle (Advent through Eastertide, per the post-Pentecost sweep's own "not done" list above)
remains open and is unrelated to this particular gap.

**Ordering bug found and fixed same day**: pooling a `temporalId`'s own file (Nocturn 3 only,
for a late post-Pentecost Sunday) alongside a separately-sourced `month-week-<id>` file
(Nocturn 2 only) put the Nocturn 3 reading ahead of the Nocturn 2 reading in the pool — the two
sources were being concatenated in `nocturnReadingIds` set order, not grouped by each reading's
own `nocturn` tag, so `distributeIntoNocturns`'s front-light positional slotting (1st pool
reading -> Nocturn 1, 2nd -> Nocturn 2, rest -> Nocturn 3) put the Nocturn-3-tagged reading into
the Nocturn 2 slot and vice versa. Confirmed live on the date that surfaced the whole month-week
gap in the first place (2026-08-23, `post-pentecost-13` + `month-week-084`): St. Gregory's
Moralia commentary (tagged `nocturn: 2`) was showing up after St. Augustine's homily (tagged
`nocturn: 3`) instead of before it. Fixed in `buildReadingPool` (`hours/matins.ts`) by bucketing
every contributing id's readings by `reading.nocturn` and emitting the buckets in ascending
numeric order — not hardcoded to `[2, 3]`, since Ember days' own nocturn-readings files use
`nocturn: 1` for their single-nocturn structure, and an initial `[2, 3]`-only version of this
fix silently dropped those (caught by `tests/calendar/ember-days.test.ts` failing). Within each
bucket, id order (and each file's own reading order) is preserved, matching this pool's usual
priority rule. `npm test` green at 460 tests — no fixture changes needed, since every
previously-authored id's own file already had its readings correctly ordered internally; only
the cross-source case this pass introduced was actually broken.

### Nocturn-readings responsory marker fix — done (2026-08-23)

Every `data/propers/nocturn-readings/*.yml` responsory with a `responsory:` field (53 files:
all 24 `post-pentecost-*`, all 20 `month-week-*`, the 6 Ember-day files, plus `st-monica`,
`st-marcellus-i`, `apparition-of-st-michael`) started straight into the Latin/English text with
no leading marker, unlike `data/propers/octave-readings/*.yml`'s own convention of an explicit
`℟.`/`℣.` at the start of each line. User flagged this via the Augustine reading on
2026-08-23's date and explicitly chose to fix it in content, not by injecting a marker at
render time. Given each file already uses plain `V.`/`R.` (not the `℣.`/`℟.` glyphs) for the
versicle and the mid-responsory repeat, `R. ` was prepended to just the opening `la`/`en` text
of each responsory to match that same plain-letter convention already used internally by these
same files — not a switch to octave-readings' Unicode-glyph convention, which remains a
separate, still-open reconciliation between the two stores if it's ever wanted. Mechanical,
scripted edit (53 files, idempotent — skips any responsory already starting `R. `), verified no
unexpected format by having the script flag anything it couldn't parse (none found). `npm test`
(460), `tsc`, and `npm run build` all still green — pure content edit, no code/mechanism change.

### Pre-1955-but-post-1617 saints' Matins readings — done (2026-08)

Resolved the former "mechanism gap" above into a concrete, scoped content pass — no new
mechanism was actually needed, since `nocturn-readings.ts`/`buildReadingPool` already accept
content from any source once authored (see the reading-pool redesign above); this was a
sourcing gap, not a design gap.

**Method**: read the reference engine's own files directly on disk (`../divinum-officium-reference`),
not live queries — `web/www/horas/Latin/Sancti/MM-DD.txt` is the Tridentine track (Sancti, no
`M`), `web/www/horas/Latin/SanctiM/MM-DD*.txt` the Monastic track. A saint with a `Sancti/`
file but no `SanctiM/` file at all (any suffix) never existed in Monastic 1617, confirming the
pre-1955-but-post-1617 case. **Critical wrinkle found and now memory'd** (see memory
`vu-reference-engine-file-layout`, `vu-matins-pre1955-saints`): the plain, unsuffixed
`Sancti/MM-DD.txt` is very often the *wrong* saint — the general/Tridentine calendar keeps
piling later canonizations onto the same date slot in that file, displacing the older saint
into a suffixed sibling file (`01-23o.txt`, `03-04cc.txt`, etc.) or, in one case (St. Paul the
Hermit), an entirely different date. Every one of the ~52 candidates was individually verified
by `[Officium]` match, not assumed from the filename.

**Scope, computed against `sanctoral-calendar.yml`'s 204 saint-day entries**: 52 pre-1955-only
candidates (`Sancti/` present, no `SanctiM/` at all, any suffix); of these, 43 had real proper
lesson content to author, 9 turned out to be pure Common-of-X with no proper text in the
source at all (`st-lucius-i`, `vigil-of-st-james`, `st-christina`, `vigil-of-st-bartholomew`,
`ss-twelve-brothers`, `vigil-of-ss-simon-and-jude`, `st-saturninus`, `st-sabbas`,
`vigil-of-st-thomas` — correctly left unauthored, not a gap). All 43 authored as
`src/data/propers/nocturn-readings/<id>.yml`, same shape and combine-where-sensible rule as
the two original proof dates. A few readings reuse already-verified text from
`post-pentecost-15.yml` where the source file itself cross-references another day's lessons
(e.g. `st-monica`'s Nocturn 3 Gospel, the same Luke 7:11-16 pericope). A handful of bare
Gospel pericopes were left unauthored where the source gives only an incipit ("Et réliqua")
and no full text was available to transcribe (flagged per-file, not silently dropped) — same
deferral as the bulk Vulgate/Douay-Rheims import. One saint (`st-anicetus`) has an
English-source transcription bug (Latin duplicated under the English heading) — this
project's own translation used instead, marked `draft`.

`npm test` (365 tests, all passing — none of the new saints are in a test table yet, same as
the bible-plan bulk conversion) and `npm run build`/`tsc --noEmit` both pass.

**Not done in this pass**: the larger pool of saints that *do* have a Monastic 1617
(`SanctiM/`) track — same authoring shape, separate/older backlog item (see priority §2 above).

### SanctiM (Monastic 1617)-track saints' Matins readings — done, all 12 months (2026-08)

The complementary pool to the pre-1955 pass above: saints with a `SanctiM/MM-DD*.txt` file.
**Method finding, worth recording**: a `SanctiM/MM-DD.txt` file almost always turns out to be
a thin *rule override* (`@Sancti/MM-DD` plus tweaks to psalmody/lesson-count/antiphons), not
independent content — when it doesn't override the Nocturn 2/3 `[LectioN]` sections, that
saint's Monastic-office vita reading is **the same text as the Tridentine `Sancti/MM-DD.txt`
file**, sourced exactly as in the pre-1955 pass above. Only a real Monastic-only saint (one
with no `Sancti/MM-DD*.txt` at all, e.g. `st-maurus-abbot`, who isn't on the general
Tridentine calendar) needs content read directly out of the `SanctiM/` file itself.

**A second wrinkle, on top of the pre-1955 pass's "don't trust the plain file" one**: the
"does a `SanctiM/MM-DD*.txt` exist for this date" scope check has to be done **per saint, not
per date** — a shared calendar date (two ids mapped to the same `MM-DD`) can have a `SanctiM`
file for one saint's own Tridentine slot while the other saint has no Monastic office at all.
Naively globbing "any `SanctiM/MM-DD*.txt` exists for this date" over-counts: found this by
hand for Jan 18 (`chair-of-st-peter-at-rome` has `SanctiM/01-18.txt`; `st-prisca`, sharing the
date, does not — she was actually a missed pre-1955-pool candidate) and Jan 14
(`st-hilary-of-poitiers` has `SanctiM/01-14.txt`; `st-felix-presbyter`, sharing the date via
`Sancti/01-14cc.txt`, does not — likewise a missed pre-1955-pool candidate). Both were authored
this pass using the pre-1955 pattern instead. **The 52/117 pre-1955/SanctiM split recorded
above is therefore an undercount on the pre-1955 side** by at least these two; the true
per-saint (not per-date) recomputation used for this pass found **152 ids with a genuine,
correctly-suffix-matched `SanctiM` file** — see memory `vu-matins-pre1955-saints` for the
running correction note.

**A third finding**: a `Sancti`/`SanctiM` date landing inside the Christmas or Epiphany octave
(e.g. `st-telesphorus` Jan 5, `st-hyginus` Jan 11 — both early popes commemorated on octave
days) can have its entire Nocturn 2/3 given over to octave-of-the-season homiletic content
(a Nativity/Epiphany sermon), with **no proper vita text for the saint at all** in either
track. Correctly left unauthored for these two — not a gap, same category as the pre-1955
pass's 9 pure-Common exceptions, just a different underlying reason (octave-absorbed, not
Common-of-X).

**January done (11 saints, 2 left unauthored per the octave finding above)**:
`st-hilary-of-poitiers`, `st-felix-presbyter`, `st-maurus-abbot` (Monastic-only; no English
source translation exists in the reference engine at all for this file, so `en` is this
project's own translation, marked `draft`), `st-anthony-abbot`, `chair-of-st-peter-at-rome`
(Nocturn 2 patristic + Nocturn 3 genuine Gospel-and-homily, both authored),
`st-prisca`, `ss-fabian-and-sebastian`, `st-agnes`, `ss-vincent-and-anastasius`,
`conversion-of-st-paul`, `st-john-chrysostom`. `st-telesphorus`, `st-hyginus` (octave-absorbed,
see above) were originally left unauthored here — **since authored, see "Octave-absorbed and
Vigil-only readings" below.** `npm test`/`tsc --noEmit` both pass; none of these are in a test
table yet, same as the pre-1955 pass.

**February done (8 saints, 1 left unauthored as Common-only, 1 vigil left unauthored as
Common-only)**: `st-ignatius-of-antioch` (Nocturn 2 patristic vita + Nocturn 3 homily-only,
bare Gospel incipit not transcribed per convention), `st-agatha`, `st-dorothea` (turned out to
be a missed pre-1955-pool candidate, not a real SanctiM case — the SanctiM file for this date
belongs to St. Titus, who shares Feb 6 and isn't on vu's own calendar), `st-romuald`,
`st-scholastica`, `st-simeon` (also a missed pre-1955-pool candidate — the SanctiM file at this
date is unrelated content for St. Bernadette Soubirous, not Simeon),
`chair-of-st-peter-at-antioch` (Nocturn 2 patristic + Nocturn 3 homily-only), `st-matthias`
(Nocturn 3 homily-only; Nocturn 2's own Lectio4-6 falls back to Common of an Apostle, not
proper text). Left unauthored: `st-valentine` (Lectio93 itself is `@Commune/C2:Lectio4` —
Common-only, no proper text in the source at all) and `vigil-of-st-matthias` (Vigilia rank,
`ex C1v`, no Lectio content at all — Common-only). `npm test`/`tsc --noEmit` both pass.

**A fourth SanctiM-pass finding, from this month**: a bare Gospel incipit ("Et réliqua") with
a genuine patristic homily following it (`st-ignatius-of-antioch`'s Nocturn 3,
`chair-of-st-peter-at-antioch`'s Nocturn 3, `st-matthias`'s Nocturn 3) is now authored as the
homily alone, `isGospel: false`, not combined with the bare incipit line and not marked
`isGospel: true` — cleaner than `chair-of-st-peter-at-rome`'s January file, which included the
incipit line under `isGospel: true` despite not having the full pericope text. That January
file is left as-is (not wrong, just an earlier, less clean judgment call) — this convention
applies going forward.

**March done (6/6 saints)**: `st-thomas-aquinas`, `st-frances-of-rome`, `forty-holy-martyrs`,
`st-gregory-the-great` (Lectio7-9's Gospel+homily is generic Chair-of-Peter content copied
verbatim into this file, unrelated to Gregory — same lessons as
`chair-of-st-peter-at-rome.yml`'s own sourcing, not proper to this feast, so not authored
here), `st-joseph`, `st-benedict` (Duplex I class patronal feast — the richest SanctiM case yet:
Nocturn 2 sourced from `Sancti/03-21.txt`'s own vita per the usual pattern, but Nocturn 3 is
**genuinely Monastic-only content not found in `Sancti/` at all** — a St. Peter Damian sermon
preached specifically for Benedict's own feast; no `English/SanctiM/03-21.txt` exists in the
reference engine, so `en` for that reading is this project's own translation, marked `draft`).
`npm test`/`tsc --noEmit` both pass.

**April done (9/9 saints)**: `st-leo-i`, `ss-tiburtius-valerian-and-maximus` (missed
pre-1955-pool candidate — the date's only SanctiM file belongs to St. Justin, who shares Apr 14
and isn't sourced this way), `st-george` (a reference-engine data bug: `Sancti/04-23.txt`'s own
`[Lectio94]` Latin section actually contains English prose — same kind of bug as `st-anicetus`
in the pre-1955 pass — so `la` here is this project's own translation, marked `draft`;
`en` is the source text, genuinely present just filed under the wrong heading, `verified`),
`st-mark`, `ss-cletus-and-marcellinus`, `st-vitalis`, `st-robert` (genuinely Monastic-only, no
`Sancti/04-29*.txt` at all; no English source translation exists either, so `en` is this
project's own translation, `draft`), `st-catherine-of-siena` and `st-peter-martyr` (a newly
found **date-swap wrinkle**: this app's calendar has Catherine on Apr 29 and Peter Martyr on
Apr 30, the *opposite* of the general Tridentine calendar's own Apr 29/Apr 30 pair — both
sourced from their real, off-date Tridentine files, the same off-date pattern as St. Paul the
Hermit in the pre-1955 pass; Peter Martyr in particular has no genuine SanctiM match at either
date, so also a missed pre-1955-pool candidate). `npm test`/`tsc --noEmit` both pass.

**May done (12/13 saints, 1 left unauthored as Common-only)**: `ss-philip-and-james` (Nocturn 3
has a *full* Gospel pericope, John 14:1-13, not a bare incipit — the first case since
`chair-of-st-peter-at-rome` in January where `isGospel: true` was actually warranted),
`st-athanasius`, `ss-alexander-eventius-theodulus-and-juvenal` (missed pre-1955-pool candidate —
SanctiM's own suffix-matched file just narrows the joint feast, same pattern as
`ss-vincent-and-anastasius`/`ss-cletus-and-marcellinus`), `st-john-before-the-latin-gate`,
`st-gregory-nazianzen`, `ss-nereus-achilleus-domitilla-and-pancras`, `st-boniface` (missed
pre-1955-pool candidate — the date's only SanctiM file is a different, later saint), `st-ubald`
(same), `st-peter-celestine`, `st-pudentiana` (missed pre-1955-pool candidate — shares its date
with Peter Celestine, whose SanctiM file the naive per-date check matched instead), `st-urban-i`
and `st-john-i` (both missed pre-1955-pool candidates via the "plain file is the wrong,
later saint" wrinkle). Left unauthored: `st-romanus-abbot` — turns out to have a real
`SanctiM/05-22M.txt` file after all (resolving the earlier "neither file" mystery from the
pre-1955 pass), but it carries no Lectio content at all, only `vide C5b` (Common of a
Confessor) — Common-only, not a gap. `npm test`/`tsc --noEmit` both pass.

**June done (12/13 saints, 1 left unauthored as no-proper-content)**:
`ss-marcellinus-peter-and-erasmus`, `ss-primus-and-felicianus`, `st-barnabas`,
`st-basil-the-great`, `ss-vitus-modestus-and-crescentia`, `ss-mark-and-marcellian` (missed
pre-1955-pool candidate), `nativity-of-st-john-the-baptist`, `st-william-abbot` (a genuine
Monastic-only vita exists too, but its content is essentially identical to the Tridentine
track's own account and only the Tridentine track has an English translation, so sourced from
there instead), `ss-john-and-paul`, `st-leo-ii` (off-date, like St. Paul the Hermit in the
pre-1955 pass — the general calendar keeps him on Jul 3, not Jun 28), `ss-peter-and-paul`
(Duplex I class), `commemoration-of-st-paul`. `vigil-of-st-john-the-baptist` was originally left
unauthored here as "only a Nocturn 1, no content" — **wrong: `Lectio1-3` turned out to be real
homiletic content, just filling the Vigil's simpler single-nocturn slot; since authored, see
"Octave-absorbed and Vigil-only readings" below.** `npm test`/`tsc --noEmit` both pass.

**July done (13/18 saints, 5 left unauthored as Common-only)**: `ss-seven-brothers-rufina-and-secunda`,
`st-pius-i`, `st-john-gualbert`, `st-anacletus`, `st-alexius`, `ss-symphorosa-and-seven-sons`
(the last four missed pre-1955-pool candidates — see the ninth finding below), `st-bonaventure`
(a reference-engine translation-misalignment bug — its English section boundaries don't line up
with the Latin, so `en` here is this project's own translation, `draft`), `st-mary-magdalene`,
`st-apollinaris`, `st-james-the-greater`, `st-anne`, `st-martha`, `ss-abdon-and-sennen`. Left
unauthored (Common-only, no proper text at all in the source): `ss-nabor-and-felix`,
`st-margaret-of-antioch`, `st-liborius`, `st-christopher`, `ss-felix-simplicius-faustinus-and-beatrix`.
`npm test`/`tsc --noEmit` both pass.

**A ninth finding, from July**: the Monastic calendar keeps a full **Octave of St. Benedict**
(Jul 11-18, since his own feast is Mar 21 but his Monastic-only "Transitus" observance sits in
July) — every SanctiM file at these dates that isn't a saint's own real match turns out to be
"Nth day within the Octave of St. Benedict" content, not a genuine candidate mismatch to
puzzle over. Recognize this pattern (`In Octava S. P. N. Benedicti Abbatis`) and move on
quickly rather than re-investigating it saint by saint.

**August done (13/16 saints, 3 left unauthored as no-proper-content)**: `st-peter-ad-vincula`,
`ss-machabees` (missed pre-1955-pool candidate), `st-stephen-i` (missed pre-1955-pool, "plain
file is a different, later saint" trap), `st-dominic`, `ss-cyriacus-largus-and-smaragdus`,
`st-clare` (missed pre-1955-pool candidate), `st-lawrence` (Duplex I class), `st-bernard`,
`st-bartholomew`, `st-augustine`, `decollation-of-st-john-baptist`, `st-sabina` (missed
pre-1955-pool candidate), and `assumption` — the Duplex I class feast of the B.V.M. itself,
confirming this app's calendar does map the Assumption despite the parent CLAUDE.md's older
note that fixed BVM/Lord feasts were "not modeled by this app at all yet"; Nocturn 1's own
scripture (Song of Songs) left to the bible-plan store per this project's usual split, only
Nocturn 2/3 patristic/Gospel content authored here. Left unauthored (no proper content at all
in the source, same category as prior Common-only/Vigil exceptions): `ss-nabor-and-felix`-style
`st-romanus` and `st-hermes` (both Common-only). `vigil-of-st-lawrence` was originally left
unauthored here as "Nocturn 1 only, no content" — **wrong, same correction as
`vigil-of-st-john-the-baptist`; since authored, see "Octave-absorbed and Vigil-only readings"
below.** `npm test`/`tsc --noEmit` both pass.

**September done (9/9 saints, full month)**: `nativity-bvm`, `st-hadrian` (missed pre-1955-pool
candidate), `st-matthew`, `ss-maurice-and-companions` (missed pre-1955-pool), `st-linus`,
`st-thecla` (missed pre-1955-pool), `ss-cosmas-and-damian`, `st-michael` (Dedication of St.
Michael the Archangel — its SanctiM file structurally borrows from his own separate May 8
Apparition file but the actual Lectio content resolves back to this day's own Sancti file),
`st-jerome`. `npm test`/`tsc --noEmit` both pass.

**October done (9/13 saints, 4 left unauthored as Common-only or no-content)**: `st-remigius`,
`st-francis`, `ss-placid-and-companions`, `st-marcus` (missed pre-1955-pool candidate),
`st-callistus`, `st-luke`, `st-hilarion`, `ss-chrysanthus-and-daria`, `ss-simon-and-jude`. Left
unauthored: `st-justina` and `st-gall` (both genuinely Monastic-only — no `Sancti/` file at all
— but their own `SanctiM` files are themselves Common-only, `vide C6`/`vide C5b`, no proper
Lectio content to source), `ss-ursula-and-companions` (Common-only, no proper content — matches
this file's own earlier flag in the "individually-flagged content gaps" section), and
`vigil-of-all-saints` (re-checked per "Octave-absorbed and Vigil-only readings" below: its
`Lectio1-3` are genuinely a bare `@Commune/C3:Lectio7-9 in 2 loco` rubric, not real content like
the other two vigils below turned out to be — correctly still unauthored). `npm test`/`tsc
--noEmit` both pass.

**November done (14/16 saints, 2 left unauthored as Common-only)**: `all-saints`,
`st-martin-of-tours`, `st-menna` (missed pre-1955-pool), `st-martin-i` (off-date, like St. Paul
the Hermit — general calendar keeps him Nov 12, this app's own Nov 12 goes to St. Emilian),
`all-saints-of-the-benedictine-order` (genuinely Monastic-only, and unusually one of the few
such cases with both rich content *and* an available English translation),
`dedication-of-the-basilicas-of-ss-peter-and-paul` (missed pre-1955-pool; one sentence of the
source's own English translation was left untranslated, so `en` for that reading is `draft`),
`st-pontian` (missed pre-1955-pool), `st-cecilia`, `st-clement`, `st-felicitas` (missed
pre-1955-pool), `st-chrysogonus` (missed pre-1955-pool), `st-silvester-abbot`,
`st-peter-of-alexandria` (missed pre-1955-pool), `st-andrew`. Left unauthored (Common-only, no
proper text at all in the source): `st-emilian`, `st-odo` — both genuinely Monastic-only but
with no Lectio content in their own SanctiM files, the same wrinkle as `st-justina`/`st-gall`
in October. `npm test`/`tsc --noEmit` both pass.

**December done (13/15 saints, 2 left unauthored as Common-only) — final month, pass
complete**: `st-bibiana`, `st-nicholas`, `st-ambrose`, `immaculate-conception` (two genuinely
distinct Nocturn 2 sources kept as separate readings — St. Jerome's sermon, and the Acts of Pius
IX defining the dogma in 1854), `st-damasus`, `st-lucy`, `st-thomas-apostle`,
`st-stephen-protomartyr`, `st-john-apostle`, `holy-innocents`, `st-thomas-becket`,
`st-silvester-i`, `st-eusebius-of-vercelli` (off-date, missed pre-1955-pool candidate — vu
places him Dec 15 via Monastic 1617's own same-day redirect, but his real proper content is at
the general Tridentine calendar's Dec 16; no SanctiM file exists for him at either date). Left
unauthored (Common-only, no proper text at all in the source): `st-barbara` (correct file
`Sancti/12-04o.txt`) and `st-melchiades` (correct file `Sancti/12-10o.txt` — this also corrects
an earlier wrong "neither Sancti nor SanctiM file" claim about him in memory
`vu-matins-pre1955-saints`, the same "verify by opening the file, not just checking existence"
lesson as November's St. Romanus Abbot correction). Adding `st-eusebius-of-vercelli`'s content
surfaced a 3rd Matins lesson on the fixed Advent III Tuesday proof date (`tests/hours/
matins.test.ts`, Dec 15 falls under his commemoration) — updated that test's expected lesson
count from 2 to 3, not a regression: exactly the "commemorations get their own reading" design
working as intended once real content exists to surface. `npm test` (365 tests) and `tsc
--noEmit` both pass.

**Pass complete**: all 12 months (January-December) of the SanctiM-track saints' Matins
readings are now done — 133 saints authored across the whole year, the rest correctly left
unauthored as Common-only/no-content cases. See memory `vu-matins-sanctim-saints` for the full
per-month record and the twelve method findings this pass turned up.

### Octave-absorbed and Vigil-only readings — corrected policy, 4 saints authored (2026-08)

**Direct instruction, correcting this pass's own earlier judgment call**: "octave-absorbed"
(the day's Nocturn 2/3 given over to a Nativity/Epiphany homily instead of the saint's own
vita) and "Vigil, Nocturn 1 only" were both being treated as "no content, correctly left
unauthored" — wrong. This app pools readings first and figures out how/where to display them
later (see `hours/matins.ts`'s own `buildReadingPool` header); real, sourceable text at a
saint's own calendar slot should go into that pool even when it isn't proper to that particular
saint, rather than being withheld because it came from "a different rubric variant." Four
saints previously left unauthored for these two reasons are now authored, flagged in each
file's own header as not-proper-to-the-saint:

- `st-telesphorus` (Jan 5) — St. Augustine's Christmas-octave sermon (Nocturn 2) + St. Jerome's
  homily on Matt 2:19-23 (Nocturn 3), neither proper to Telesphorus himself.
- `st-hyginus` (Jan 11) — St. Fulgentius's Epiphany sermon (Nocturn 2) + St. Ambrose's homily on
  Matt 2:1-12 (Nocturn 3), neither proper to Hyginus himself.
- `vigil-of-st-john-the-baptist` (Jun 23) — St. Ambrose's homily on Zachary and Elisabeth
  (Luke 1:5-17), authored from `Lectio1-3`, the Vigil's own single 3-lesson nocturn (not the
  usual `Lectio4-9` Nocturn 2/3 slot a full feast uses).
- `vigil-of-st-lawrence` (Aug 9) — Pope St. Gregory the Great's homily on self-denial
  (Matt 16:24-27), same `Lectio1-3` single-nocturn source.

`vigil-of-all-saints` (Oct 31) was re-checked under this same corrected policy and confirmed
still genuinely contentless — its `Lectio1-3` are a bare `@Commune/C3:Lectio7-9 in 2 loco`
rubric with no proper text at all, unlike the two Vigils above. Not every "left unauthored"
case from this pass needs revisiting — only the octave-absorbed and Vigil-only categories were
affected by this policy correction; genuine Common-only cases (a bare `@Commune` rubric with no
Lectio text anywhere) are unaffected and remain correctly unauthored.

Adding these introduced no new lesson-count regressions in `tests/hours/matins.test.ts` (none
of its fixed proof dates fall on Jan 5/11 or Jun 23/Aug 9). `npm test` (365 tests) and `tsc
--noEmit` both pass.

### Octave-readings audit — two real gaps found and closed, scope of the rest recorded (2026-08-28)

Prompted by the user pointing out that the reference source keys octave-day content by
**calendar date** (e.g. `08-11oct.txt` through `08-16oct.txt` for St. Lawrence's octave), not
grouped by octave-day-number the way `data/propers/octave-readings/<feast>-octave-day-N.yml` is
— a genuinely different question from the saint-level sweep above, not covered by it. Audited
every fixed-date octave (`st-lawrence`, `assumption`, `all-saints`, `immaculate-conception`,
`nativity-bvm`) for missing days: most gaps in the existing files (IC days 4/6, Nativity-BVM day
7, Lawrence days 6/7) turned out to be days where the source's own `[Rank]` says `vide Sancti/
MM-08` with **no Lectio override at all** — genuinely no distinct content to author, matching the
existing precedent already set by St. Lawrence's own correctly-unauthored days. Two, however,
were real gaps with genuine, un-authored Lectio content:
- `immaculate-conception-octave-day-8.yml` (Dec 15, the octave's own closing/"In Octava" day) —
  Pope Pius IX's dogmatic bull (a different excerpt than any other day in the octave) + a
  homily of St. Epiphanius of Salamis on the BVM (Lectio7-9, bare Gospel incipit dropped per the
  usual rule). Both languages present and verified in the source.
- `nativity-bvm-octave-day-3.yml` (Sep 10) — St. Jerome's homily on Matthew's genealogy
  (Matt 1:1-17, bare incipit dropped). No usable English exists in the reference engine for this
  file — `English/Sancti/09-10bmv.txt` is a verbatim Latin duplicate, the same reference-engine
  data-bug class as `st-anicetus`/`st-george` — so `en` is this project's own translation,
  `draft`.

Both live-resolved correctly (`resolveOrdo` on 2026-12-15 and 2026-09-10) and `tsc --noEmit`/
`npm test` (522 tests) both pass.

### Christmas-week saint octaves (Stephen, John, Holy Innocents) — done (2026-08-28)

`st-stephen-protomartyr`, `st-john-apostle`, `holy-innocents` (Dec 26/27/28's own 8-day octaves)
had zero `octave-readings` files. Their date range overlaps Christmas's own already-authored
Vespers/Lauds `octave-commemoration` blocks (see "Octave content — done" above) — a different
hour/mechanism, not a substitute for Matins content. Checked every suffix file at every date in
the three overlapping ranges (Dec 27–Jan 4): the intervening days (2-7) carry **no Lectio
override at all** in the source — every suffix file found (`12-27o`, `12-28o`/`12-28r`,
`12-29o`/`12-29r`, `12-31o`/`12-31r`; Dec 30 has no suffix file whatsoever) is a bare
`@Sancti/MM-DD` alias, correctly left unauthored, same "vide-only, not a gap" pattern as the
fixed-date octaves above — these three minor commemorations simply have no standing against
Christmas week's own dominant office on those days. Only each octave's own closing/"In Octava"
day carries real content: `st-stephen-protomartyr-octave-day-8` (Jan 2 — St. Augustine's own
sermon on Stephen + St. Jerome on Matt 23:34-35), `st-john-apostle-octave-day-8` (Jan 3 —
St. Augustine's Tractate 36 on John, the "eagle" imagery + St. John Chrysostom on John 21:20-24),
`holy-innocents-octave-day-8` (Jan 4 — St. Augustine's own first sermon for Childermas +
St. John Chrysostom on Matt 2:13-18). Holy Innocents' own Lectio4 closing sentence in the
source's English is garbled into an incoherent fragment (a legal-term wordplay on "pledges,"
"Pígnora sunt, non crédita, sed creáta...") — reconstructed to match the Latin, `en` marked
`draft` for that file. `tsc --noEmit` and `npm test` (522 tests) both pass; live-verified via
`activeOctavesFor`/`resolveOrdo` on 2027-01-02/03/04, confirming all three readings surface
correctly (and don't collide) even on dates where two or three of these octaves are
simultaneously active. This closes out the last open item from the "octave-readings audit"
above — no octave-reading gaps remain open anywhere in the calendar.

### Movable-feast octaves (Ascension, Corpus Christi, Sacred Heart) — done (2026-08-28)

Direct correction from the user: movable-feast octave content lives in the reference engine's
`Tempora/`/`TemporaM/` directories, keyed by liturgical offset from Easter (e.g. `Pasc5-4.txt` for
Ascension, `Pent02-5.txt` for Sacred Heart — each feast's own record in `data/calendar/
temporal-feasts/*.yml` already cites its own file), not by calendar date — no representative year
needs picking, contrary to what this file previously assumed. All three octaves' full 7 non-feast
days (day 2 through the closing day 8) turned out to have genuine, distinct Lectio4-9 content —
unlike the fixed-date octaves audited above, none of these turned out to be bare `vide`-only days.
21 files authored: `ascension-octave-day-2` through `-8` (St. Leo/St. Augustine/St. John
Chrysostom/St. Gregory of Nyssa sermons + Pope St. Gregory the Great's own homily on Mark
16:14-20, reused across all 7 days), `corpus-christi-octave-day-2/3/5/6/7/8` (St. Thomas Aquinas/
St. John Chrysostom/St. Ambrose/St. Cyril of Jerusalem/St. Cyril of Alexandria/St. Cyprian +
St. Augustine's own Tractates on John 6, reused across all days), `sacred-heart-octave-day-2/4/5/
6/7/8` (excerpts from Pius XI's "Miserentissimus Redemptor" + homilies on John 19:31-37 by
St. John Chrysostom/St. Lawrence Justinian/St. Bernardine of Siena/St. Peter Canisius/St. Bernard/
St. Augustine).

**One real wrinkle found and fixed**: each octave's own day-4-or-day-3 (the Sunday falling within
it) turned out to duplicate content *already authored* under a different id in the earlier
temporal-cycle Sunday sweep (see "Temporal-cycle nocturn-readings sweep" below) —
`corpus-christi-octave-day-4` duplicated `post-pentecost-02` (both source `Tempora/Pent02-0.txt`),
`sacred-heart-octave-day-3` duplicated `post-pentecost-03` (`Tempora/Pent03-0.txt`) — because a
numbered post-Pentecost Sunday and the Sunday within one of these two octaves are, by construction
(both anchored to Easter with fixed offsets), always the same calendar day every year, not a
one-off 2026 coincidence. Both duplicate files were deleted after live-verifying the reading
already surfaced correctly (and only once) via the existing temporal id. Ascension's own Sunday
(`ascension-octave-day-4`, `Pasc6-0`) has no such collision — Paschaltide's own temporal-cycle
sweep hasn't been done yet, so no duplicate content exists there yet. If the Paschaltide sweep is
picked up later, re-check `ascension-octave-day-4` against whatever gets authored for that Sunday
before assuming it's still needed. `tsc --noEmit` and `npm test` (522 tests) both pass;
live-verified via `activeOctavesFor`/`resolveOrdo` for representative dates in 2026 across all
three octaves, confirming correct day-numbering and no residual duplicate readings in the pool.

### The Patronage of St. Joseph — new movable feast + octave, added (2026-08-28)

A real gap, not just a content pass: this feast (Duplex I classis, instituted by Pius IX 1847)
didn't exist anywhere in the calendar at all. Per direct instruction, dated per the **older,
pre-1955 Wednesday reckoning** — the Wednesday after the 2nd Sunday after Easter (Easter+17) —
rather than the 1913 Pius X reform's later placement on the 3rd Sunday after Easter itself (the
reference engine carries both: the primary, unsuffixed `Tempora/Pasc2-3.txt` is the Wednesday
dating used here; `Pasc3-0t.txt`, a bare alias back to `Pasc2-3`, is the Sunday dating, not used).
Added the mechanism — `data/calendar/temporal-feasts/patronage-of-st-joseph.yml` (rank, Easter+17
anchor, `wins: duplex` octave) plus a new `[17, 'patronage-of-st-joseph']` entry in
`calendar/temporal-feasts.ts`'s `EASTER_OFFSET_STARTS` table (the one piece of *code*, not data,
this needed) — then authored the feast day's own Matins reading (`nocturn-readings/
patronage-of-st-joseph.yml`, St. Bernardine of Siena + St. Augustine on the two genealogies) and
its full movable octave: days 2/3/4 (Thu/Fri/Sat, real distinct content each day — Bernardine,
Chrysostom, Bernard, Ambrose, John Damascene, Augustine sermons/homilies rotating through), day 6/
7/8 (Mon/Tue/Wed, likewise), and day 5 (the Sunday, `Pasc3-0`) correctly left unauthored — the
source itself has no override there at all, unlike Corpus Christi/Sacred Heart's own octaves.

**One real test collision found and fixed**: `tests/calendar/may-sanctoral.test.ts`'s existing
2038-05-19 proof date (St. Pudentiana losing to St. Peter Celestine) broke, because 2038-05-19
happens to be this new octave's own closing day, and Celestine's `duplex` rank exactly ties the
octave's own closing-day threshold — the app's own already-documented tie-break
(`calendar/index.ts`'s `applyOctaves`) correctly favors the octave on a tie, displacing Celestine
into a commemoration that year. Not a bug; moved the proof date to 2033 (clear of any octave) to
keep testing what it was meant to test. Also bumped `vite.config.ts`'s PWA
`maximumFileSizeToCacheInBytes` from 6 MiB to 8 MiB — this content pull pushed the main bundle
(6.32 MB) past the previous cap; same deliberate-not-bloat headroom bump as every previous time
this has happened (see that file's own comment). `tsc --noEmit`, `npm test` (522 tests), and
`npm run build` all pass; live-verified via `resolveDay`/`activeOctavesFor`/`resolveOrdo` that the
feast wins outright on its Wednesday, the octave stays active through the week without disturbing
Sundays or stronger saints, and every reading surfaces correctly and exactly once.

### St. Felix I (May 30) — one missed pre-1955-pool candidate found, authored (2026-08-28)

Re-auditing the full pre-1955/SanctiM two-pass sweep (diffing every id in
`sanctoral-calendar.yml` against `data/propers/nocturn-readings/*.yml`, then checking every
"missing" id's suffix-matched `Sancti/SanctiM` files by hand) turned up one genuine gap neither
pass's own per-month log had caught: `st-felix-i` (May 30). It fits the pre-1955-pool criterion
exactly — a real `Sancti/05-30.txt` `[Lectio93]` vita (both languages present), no
`SanctiM/05-30.txt` of its own (the only SanctiM file at that date, `05-30b.txt`, belongs to St.
Joan of Arc, a different feast sharing the date) — but was absent from both memory records'
scoping passes. Authored the same single-lesson shape as `st-pontian.yml` (a near-identical
simplex Pope-Martyr case). Every other id absent from `nocturn-readings/` was re-confirmed
against the existing "left unauthored" lists in both memory files and TODO.md's own log above —
all genuinely Common-only/no-content, not further gaps. `tsc --noEmit` and
`tests/hours/matins.test.ts` both pass; live-resolved on 2027-05-30 to confirm the new reading
surfaces in the pool. This closes out the sanctoral (saint-level) side of "author
patristic/hagiographic readings for every saint where one can be sourced" — the temporal-cycle
side (Advent/Christmastide/post-Epiphany/Septuagesima-Lent-Passiontide/Eastertide Nocturn 2/3
content) remains open, unstarted, see "Temporal-cycle nocturn-readings sweep" progress note.

### Aug 5 martyrology English gap, and "Our Lady of the Snows" added (2026-08)

Two separate fixes, both discovered by re-checking a claim in this file that turned out stale.

**Martyrology, Aug 5**: `src/data/martyrology/08-05.yml`'s Latin text opens with "Romæ, in
Exquíliis, Dedicátio Basílicæ sanctæ Maríæ ad Nives" (the dedication of St. Mary Major, i.e.
Our Lady of the Snows) — but the file's own English translation silently dropped the
corresponding sentence, starting straight in on the day's other martyrs instead, even though
the reference engine's own `English/Martyrologium/08-05.txt` has it ("Upon the 5th day of
August is kept the dedication of the Church of St. Mary in the Snow upon the Esquiline Hill at
Rome, in the year 366."). Restored.

**"Our Lady of the Snows" (Aug 5) added to the calendar.** `sanctoral-calendar.yml`'s own
August comment had excluded her entirely as "a fixed BVM feast, same convention as
Purification/Annunciation" — wrong, on re-reading the actual source: unlike the truly
unmodeled cases in that same comment (saints permanently subordinate to an unmodeled octave or
to the fixed Transfiguration), she isn't blocked by any of that, and this app already models
three other fixed BVM feasts identically (`assumption`, `immaculate-conception`,
`nativity-bvm`). `Sancti/08-05.txt`'s untagged `[Rank]` block gives her a real Duplex Majus
rank with real Nocturn 2/3 content (`Lectio4-6`, the miraculous-snowfall founding legend of the
basilica of St. Mary Major — both Latin and English present, plus a simplex-fallback
`[Lectio94]` retelling not separately authored). `SanctiM/08-05.txt` confirms she's on the
Monastic 1617 track too, inheriting `@Sancti/08-05` and reslicing the same Lectio4-6 into its
own 12-lesson structure — no separate Monastic-only content.

Unlike the other three BVM feasts, she has **no proper collect or Benedictus antiphon of her
own** — her `[Rule]` says `ex C11` (Common of Feasts of the BVM), and `Commune/C11.txt`'s own
`[Oratio]`/`[Ant 1]` (with full English) supplied both. First pass at this wrongly authored
that Common text as if it were her own private file via `propers:` — corrected per direct
feedback: `collect-c11.yml` and `benedictus-antiphon-common-of-the-bvm.yml` are real, reusable
Common templates now, wired in via `SaintRecord.collectCommon`/`benedictusCommon` exactly like
any other Common-only saint (e.g. St. Barbara), not a one-off duplicate. `collect-c11`'s own
text has no `{N}` name placeholder (it addresses "fámulis tuis" generically, not the saint by
name, unlike every other `collect-c*.yml`) — `collectName` is still required by
`getDayCollect`'s own lookup, so set to empty arrays, never actually substituted.

Live-checked via direct resolution (not a browser query, per
[[vu-reference-engine-file-layout]]): she wins outright on 2026-08-05 and 2027-08-05, with her
Benedictus antiphon, day collect, and Matins reading all resolving `verified`. `npm test` (365
tests) and `tsc --noEmit` both pass.

### Four more "not modeled" fixed Lord's/BVM feasts added (2026-08)

Per direct instruction ("import the sanctoral content that exists" — this app isn't
reconstructing any one calendar, and content beyond what the user's own final calendar keeps is
still worth having so others can build their own): swept every remaining "fixed Lord's/BVM
feast, not modeled" exclusion in `sanctoral-calendar.yml` the same way Our Lady of the Snows was
just fixed, rather than treating that as a one-off. All four turned out to have real, complete
source content too:

- **Finding of the Holy Cross** (May 3, Duplex II Classis) — her `[Rule]` is `proprium`: fully
  own text throughout (collect, antiphon, Nocturn 2/3). Now wins outright over
  `ss-alexander-eventius-theodulus-and-juvenal`, exactly matching what that saint's own file
  already documented as the live-verified reality ("loses the day to the Finding of the Holy
  Cross... this isn't just an artifact of the specific year checked") — her file had simply
  never been given a real competitor to lose to until now.
- **The Transfiguration of Our Lord** (Aug 6, Duplex II Classis) — same `proprium` shape, fully
  own text throughout. Its three subordinate saints (Ss. Xystus/Felicissimus/Agapitus) are
  still excluded — no longer because the Transfiguration itself is unmodeled, but because of
  the same real, separate octave/subordination-rendering gap that already blocks St. Clare et
  al.
- **The Exaltation of the Holy Cross** (Sep 14, Duplex Majus) — her `[Rule]` is
  `ex Sancti/05-03`: reuses the Finding of the Holy Cross's content wholesale. Collect/antiphon
  reuse via a cross-id `propers` pointer (`propers: finding-of-the-holy-cross`, same pattern as
  `st-michael.yml`'s `apparition-of-st-michael`) rather than duplicate files; the Matins
  nocturn-readings mechanism has no equivalent cross-id lookup (`getNocturnReadings` keys
  strictly off the saint's own id), so `exaltation-of-the-holy-cross.yml` duplicates
  `finding-of-the-holy-cross.yml`'s reading content instead.
- **Dedication of the Basilica of the Most Holy Savior** (Nov 9, i.e. the Lateran, Duplex II
  Classis) — her `[Rule]` is `ex C8` (Common of the Dedication of a Church, already modeled —
  `collect-c8.yml`/`benedictus-antiphon-common-of-a-dedication.yml` already existed from
  `dedication-of-the-basilicas-of-ss-peter-and-paul.yml`, reused directly here, not duplicated),
  but real unique `Lectio1-6(+6a)` vita content of her own. Her secondary, St. Theodore, is
  still excluded — same real octave/subordination gap as the Transfiguration's secondaries
  above (her own embedded `[Commemoratio 5]` block already carries his real collect too, ready
  to use once that mechanism exists).
- **The Presentation of the BVM** (Nov 21, Duplex Majus, itself a Monastic-1617-only addition)
  — has a real, unique `[Oratio]` of her own (unlike Our Lady of the Snows), but her
  `[Ant 1]` is a *substituted* Common reference (`@Commune/C10:Ant 2_`, with a
  feast-specific `s/Christo:.*/Christo, allelúja./` truncation applied and computed by hand,
  not shared verbatim by any other saint) — authored as her own file via `propers`, same
  "substitution makes shared-formula text effectively proper" reasoning as
  `st-lucy-collect.yml`.

**Correction to Our Lady of the Snows found during this same sweep**: her own file had no
Lectio7-9 either, and — like the Presentation of the BVM — this means she defers to
Commune/C11's own Nocturn 3 (a bare Gospel incipit, Luke 11:27-28, plus the Venerable Bede's
homily). Missed in the first pass (only her Nocturn 2 vita was authored); added now, identical
content to `presentation-of-the-bvm.yml`'s own Nocturn 3, duplicated for the same
no-cross-id-lookup reason as Exaltation/Finding of the Holy Cross above.

**One real bug caught before committing**: Finding of the Holy Cross's own collect/antiphon
files were never actually created in the first draft (only Exaltation's cross-id pointer to
them was written) — both resolved `missing` on live verification. Caught by re-checking every
new feast end-to-end via direct resolution before committing, not assumed correct from the
file list alone; fixed immediately, re-verified.

**Test updates, not regressions**: three tests had fixed proof dates that happen to collide
with the newly-added feasts, all updated to reflect the new, correct behavior rather than the
old absence: `tests/calendar/may-sanctoral.test.ts` (Finding of the Holy Cross now correctly
outranks its stub competitor instead of winning by the stub's own default), and
`tests/calendar/day-label.test.ts` (two dates — Nov 9 and 1943-11-21 — now show the new saint's
name instead of a ferial fallback label, the latter because Duplex Majus already meets this
app's own pre-existing privileged-Sunday rank threshold in `commemorations.ts`, not new
behavior). `npm test` (365 tests) and `tsc --noEmit` both pass.

### Matins psalmody overrides — mechanism built, one proof (2026-08)

Real bug found and fixed: `hours/matins.ts`'s `threeNocturns` gate (Sunday, or a Duplex+
sanctoral winner) correctly governed *nocturn count*, but the code also unconditionally reused
the literal Sunday psalmody (`matins-sunday-antiphons.yml`, Ps 20-31) as *content* for every
three-nocturn day — including a Duplex+ weekday feast, which is wrong. Confirmed live (Monastic
Tridentinum 1617, `command=prayMatutinum`) against St. Lawrence's own octave closing day
(2026-08-17, "In Octava S. Laurentii Martyris ~ Duplex"): the real psalmody is Nocturn 1 = Ps
1,2,4,5,8,10, Nocturn 2 = Ps 14,16,20,23,63,91, each with its own antiphon, plus 3 OT canticles
in Nocturn 3 (Sirach 14/15, Jeremiah 17, Sirach 31) — a genuinely different scheme from both
the real-Sunday one and the plain ferial weekday table.

Cross-checked against a second Duplex saint (St. Ignatius of Antioch, Common of a
Martyr-Bishop, no proper content of his own, 2029-02-01): Nocturn 1 identical to Lawrence's
own; Nocturn 2 nearly identical (only one of six slots differs). So the psalm *numbers* are a
shared per-Common-category pool, not unique per saint — only the antiphon text is proper to
the individual saint when authored.

Mechanism: `hours/matins-psalmody-overrides.ts` (mirrors `hours/lauds-psalmody-overrides.ts`'s
shape, but keyed differently — see its own doc comment for why per-category, not per-feast,
is the right key here) + `data/hours/matins-psalmody-overrides/*.yml`. `matins.ts`'s
three-nocturn branch now splits on `day.weekday === 'sunday'` first; a non-Sunday three-nocturn
day looks up an override by the office winner's id, falling back to the plain ferial weekday
table (redistributed into 3 nocturns instead of 1) when none is authored yet — never blank, but
honest about not yet having that saint's real proper psalmody.

**Proof authored**: St. Lawrence only (`matins-psalmody-overrides/st-lawrence.yml`) — the live-
verified content above. New scripture chapters added for the canticles:
`data/scripture/{sir-14,sir-15,jer-17,sir-31}.yml` (sparse, only the cited verses, same
convention as the pre-existing `sir-36.yml`). `matins.ts`'s canticle shape was generalized from
one scripture ref per canticle to a list of refs (`canticles: { refs: [...] }[]`), since
Lawrence's own first canticle cites two Sirach chapters under one heading in the source
("Eccli 14:22;15:3-4;15:6") — kept as one canticle, not split; `matins-sunday-antiphons.yml`
migrated to the same shape (each existing single ref wrapped in `refs: [...]`), no content
change.

**Still open** (bulk-content work, same shape as the Lauds override sweep): every other
Duplex+ Common category (Apostle, Martyr(s), Confessor, Virgin, Doctor, etc. — ~75 saints per
the earlier Matins design-memory estimate) has no override authored yet and uses the ferial-
table fallback in the meantime. A real Common-category key (vs. the per-saint key `st-lawrence`
uses now, since only proper content has been authored so far) is the natural next step once a
second category's content is authored — see `matins-psalmody-overrides.ts`'s own doc comment.

Live-verified end-to-end: `tests/hours/matins.test.ts`'s new "Duplex+ weekday-feast" describe
block (Lawrence's own psalmody resolving `verified` in both languages across all 12 psalms +
3 canticles, plus the ferial-fallback case via St. Ignatius). `npm test` (371 tests), `npm run
build`, and `tsc --noEmit` all pass.

### Matins psalmody overrides — category-keyed refactor + second proof, St. Andrew (2026-08)

Refactored `hours/matins-psalmody-overrides.ts` per its own doc comment's flagged next step:
`data/hours/matins-psalmody-overrides/` now splits into `saints/*.yml` (per-saint proper
antiphon overrides, same shape as before, St. Lawrence migrated in place unchanged) and a new
`categories/*.yml` store recording each Common category's own generic psalm-number/versicle/
canticle-ref scheme, live-verified from a saint with *no* proper antiphon of his own (so his
output is the category default outright). Two categories authored this pass:

- **`common-of-a-martyr`**: from St. Ignatius of Antioch (2029-02-01, "vide C2" — entirely
  Common, no proper text). Confirms Lawrence's one real deviation from the shared scheme worth
  remembering: his own Nocturn 2 has Ps 16 where the Common default (and Ignatius) has Ps 64 —
  a genuine proper substitution, not a transcription slip. Also confirmed (reference engine's
  `CommuneM/C2a.txt`) that **Common-of-a-Martyr-Bishop's own Matins psalmody redirects outright
  to Common-of-a-Martyr** — the two share one scheme, not two; checked all three Duplex+
  Common-of-a-Martyr-Bishop saints on the calendar (Ss. Ignatius, Callistus I, Boniface of
  Mainz) and none has a proper antiphon of his own, so this category currently has no saint
  eligible for its own `saints/` entry under the proper-only policy below — a real, not
  accidental, gap.
- **`common-of-an-apostle`**: from St. Bartholomew (2026-08-24, generic "In omnem terram"
  antiphon). Second live proof authored against it: **St. Andrew** (`saints/st-andrew.yml`,
  2026-11-30, his own day outright that year), whose psalm numbers match the category's
  generic scheme in every one of the 12 Nocturn-1/2 slots — only the antiphon text is his own
  proper narration of his call and martyrdom. New scripture chapters authored for his Nocturn 3
  canticles: `data/scripture/{isa-61,sap-3,sap-10}.yml`.

**Category files are reference/documentation content only** — not consulted by
`getMatinsPsalmodyOverride`, which still does a plain per-saint lookup exactly as before.
Direct decision from the user (2026-08): a saint without their own `saints/` entry stays on the
plain ferial-table fallback, never falls through to a category's generic antiphon text, even
though that text is real and verified — "prefer real text over missing" is deliberately *not*
extended to this case yet. The category files exist so the next proper-antiphon saint authored
in an already-proven category doesn't have to re-derive/re-verify the same shared psalm numbers
and versicles from scratch, and as the reference point for spotting a saint's real deviations
(like Lawrence's Ps 16) from noise.

Live-verified end-to-end: `tests/hours/matins.test.ts`'s new "second Duplex+ weekday-feast
proof" describe block (Andrew's own psalmody resolving `verified` in both languages across all
12 psalms + 3 canticles). `npm test` (374 tests), `npm run build`, and `tsc --noEmit` all pass.

**Still open**: ~73 remaining Duplex+ saints across the other Common categories (Confessor,
Confessor-Bishop, Virgin, Doctor, etc. — the largest population, per the earlier estimate) —
each needs its own category file (one live `prayMatutinum` query against a no-proper-antiphon
saint) before any of its own saints' proper antiphons can be authored against it.

### Matins psalmody overrides — category tier wired in as a real fallback (2026-08)

Reversed the "category files are documentation-only" decision above now that two categories
are trusted: `getMatinsPsalmodyOverride` is now a real three-tier lookup — **proper**
(`saints/*.yml`) → **Common** (`categories/*.yml`, joined via the winner's own
`SaintRecord.common` through a new `getSaintRecord` import) → **ferial fallback**
(`matins.ts`'s own plain weekday table, unchanged, now genuinely the last resort rather than
the default for every unauthored saint). This is "prefer real text over missing" applied one
tier further, once the tier itself had proof behind it. `MatinsPsalmodyOverride` and
`MatinsPsalmodyCategory` now share a `MatinsPsalmodyScheme` base (`nocturn1/2/3`) so the lookup
can return either without a cast.

Category files gained an `aliases` field: `common-of-a-martyr.yml` lists
`common-of-a-martyr-bishop`, so all three of its Duplex+ saints (Ignatius, Callistus, Boniface —
confirmed none has proper text of his own) now resolve to real, verified Common psalmody instead
of the ferial table, with zero new authoring.

`tests/hours/matins.test.ts`'s old "falls back to ferial" case (St. Ignatius, 2029-02-01) is
rewritten as a Common-tier proof (his Nocturn 2 correctly using the Common default's Ps 64, not
Lawrence's own proper Ps 16); a fresh genuine ferial-fallback proof was added (St. John Bosco,
Common of a Confessor — no category authored yet for that Common — 2026-01-31, confirmed a
clean Saturday with no Sunday collision). `npm test` (375 tests), `npm run build`, and
`tsc --noEmit` all pass.

### Matins psalmody overrides — Confessor categories authored, both tiers identical apart from versicles (2026-08)

Two more Common categories authored, following the same live-verify-a-saint-with-no-proper-
antiphon method as Martyr/Apostle:

- **`common-of-a-confessor`** (Confessor Not a Bishop, "vide C5"): from St. Francis of Paola
  (2030-04-02, confirmed clean — safely outside Holy Week/Easter octave that year, per his own
  saint-file comment). Nocturn 1 = Ps 1,2,4,5,8,10; Nocturn 2 = Ps 14,20,23,95,96,97; Nocturn 3
  = one antiphon over 3 OT canticles (Sirach 14/15, Jeremiah 17, Sirach 31 — same refs already
  authored for `common-of-a-martyr`, reused as-is).
- **`common-of-a-confessor-bishop`** ("vide C4"): from St. Basil the Great (2035-06-14, already
  live-verified clean per his own saint-file's note that his rendered Matins is byte-identical
  to this Common despite his own `[Rule]` saying "vide C4a"/Doctor). Live-verified this Common's
  psalm numbers and antiphon text are **word-for-word identical** to `common-of-a-confessor` in
  all three nocturns — the only real difference is the three nocturn-closing versicles, which
  swap plain-confessor imagery for priestly/episcopal imagery (e.g. Nocturn 3's "Tu es sacérdos
  in ætérnum * secúndum órdinem Melchísedech"). Not modeled as an `aliases` redirect (unlike
  Common-of-a-Martyr-Bishop → Common-of-a-Martyr, a genuine zero-difference redirect) because the
  versicles really do differ and the mechanism has no partial-override/diff support — kept as
  its own full duplicate file instead.

Checked every Duplex+ saint on the calendar in both Common categories (St. Aloysius Gonzaga,
St. Anthony of Padua, St. Paschal Baylon, St. John Cantius, St. Bernardine of Siena, St. Joseph
Calasanctius, St. Raymond Nonnatus, St. Anthony Mary Zaccaria for `common-of-a-confessor`; St.
Norbert, St. Andrew Corsini for `common-of-a-confessor-bishop`) — none has a proper antiphon of
his own, so (same as Common-of-a-Martyr-Bishop before it) neither category currently has a
saint eligible for its own `saints/` entry; a real, not accidental, gap, matching the existing
policy of only authoring `saints/*.yml` where a genuine proper deviation exists.

Two live saints, St. Aloysius Gonzaga (June 21) and St. Norbert (June 6), turned out **not** to
be on the Monastic Tridentinum 1617 calendar at all despite having a `SanctiM/MM-DD.txt` file —
confirmed via `Tabulae/Kalendaria/M1617.txt` (no entry) and `Tabulae/Kalendaria/M1930.txt` (only
Norbert, added there) — both were canonized well after 1617 (Aloysius 1726, though pre-1617 in
Norbert's case his 1582 canonization still postdates the M1617 calendar's fixed content). Their
`SanctiM` stub files are shared infrastructure for later Monastic versions (1930/1963), not
proof the saint was ever actually prayed under M1617 itself — a live `Pofficium.pl` query on
their own date just falls through to a plain feria. Not a blocker for authoring the Common
category files themselves (the Common content is shared infrastructure regardless of which
Monastic version's kalendar wires a given saint into it), but worth remembering next time a
"vide C" saint's own date doesn't reproduce under `version=Monastic Tridentinum 1617` — check
`Tabulae/Kalendaria/M1617.txt` before assuming the date/year picked was wrong.

`tests/hours/matins.test.ts` gained two new Common-tier proofs (Francis of Paola,
`common-of-a-confessor`; Basil the Great, `common-of-a-confessor-bishop` — same psalm numbers as
the other, checked directly). The old ferial-fallback proof (St. John Bosco, `common-of-a-
confessor`, now superseded) was swapped for Ss. Fabian and Sebastian (`common-of-several-
martyrs`, still unauthored, 2029-01-20, already live-verified clean). `npm test` (377 tests),
`tsc --noEmit` pass.

### Calendar precedence: a tie against an octave's own elevated closing day now favors the octave (2026-08)

Real bug found and fixed, same session as the Matins psalmody fix above, discovered while
checking why 2026-08-17's Matins/label looked wrong: `calendar/index.ts`'s `applyOctaves` let a
*tie* between an occurring saint's rank and an active octave's threshold go to the saint
(`isAtLeast`, `>=`). Live-verified against the reference engine that this is wrong specifically
for an octave's own elevated *closing* day: St. Hyacinth (plain Duplex, Aug 17) against St.
Lawrence's own octave closing day (also Duplex, via `closingDayRank`'s default) — the engine's
own alternate rubric block for that date is titled "Commemoratio S. Hyacinthi Confessoris",
i.e. Hyacinth is the one merely commemorated, Lawrence's elevated closing day wins outright.
Matches `collision.ts`'s own existing tie precedent for sanctoral-vs-sanctoral collisions
("ties favor the incumbent").

Fix scoped precisely to closing days, not every octave tie: a tie against an octave's
*ordinary* (non-closing) threshold still favors the occurring saint, unchanged — confirmed by
two already-verified counterexamples this app's own tests already carried (St. Thomas of
Canterbury, Dec 29, against the Christmas Octave's ordinary threshold; St. Nicholas of
Tolentino, Sep 10, against the Nativity-BVM octave's ordinary threshold) — both would have
broken under a blanket tie-flip, confirming the closing-day-specific scope is correct.
`calendar/octaves.ts`'s `ActiveOctave` gained an `isClosingDay` field to make this
distinguishable.

Two related display bugs fixed alongside it: `calendar/day-label.ts`'s `getDayLabel` dropped
every commemoration (octave or sanctoral) whenever the day's winner was a plain sanctoral
saint, and dropped every *other* active octave besides the one picked as headline even when an
octave itself was the winner — both fixed (`otherActiveOctaveNames`), gated the same
`ordinary-feria`/anchor-day rules the rest of the function already uses so Trinity Sunday's own
technical overlap with Pentecost's octave still doesn't get named ("nobody calls it that", per
the function's own pre-existing convention). Rank was also not displayed anywhere in this app's
UI at all, for any saint — added (`formatRank`), shown after the day's own winner's name, e.g.
"St. Hyacinth, Confessor (Duplex)".

Also closed a real content gap surfaced along the way: `assumption-octave-day-3.yml` (Aug 17,
the Assumption's own day 3) had never been authored — days 2 and 4-8 existed, day 3 didn't —
sourced from the reference engine's own recurring Nocturn 2 homily + the octave-wide "Vidi
speciosam" responsory already used elsewhere in her octave.

`tests/calendar/octaves.test.ts` and `tests/calendar/day-label.test.ts` updated to the
corrected behavior (not weakened) — 2026-08-17 is now a real, live demonstrable date for the
octave-vs-octave priority pick this app's mechanism already implements, rather than the
stand-in synthetic-data test that was the only option before.

### Bible-plan TSV — bulk conversion done (2026-08)

All ~369 non-blank rows of the user's own bible-plan TSV (`~/Downloads/bible-plan.tsv`, 390
lines including section headers/blanks) are now converted into `data/hours/bible-plan/*.yml`
— 366 newly generated this pass, plus the 2 proof-date files already authored (`advent-3-tuesday`,
`post-pentecost-15-sunday`). One row (Ascension Thursday) has no readings in the source and
was correctly skipped — an honest gap in the user's own plan, not a conversion miss.

Conversion was mechanical (script, not hand-authored per file): each TSV "Day" label maps to
this app's `(temporalId, weekday)` scheme via the same rules `calendar/temporal-id.ts` uses
(ferias inherit their governing Sunday's id — e.g. "Ash Wednesday" → `quinquagesima`/`wednesday`,
"Nth Sunday after Trinity" → `post-pentecost-(N+1)` per the existing display-label convention),
or to a fixed `calendarDate` for the Dec 25 - Jan 13 stretch per the reading-pool redesign
above. Each `Reading N` column becomes its own `passages` entry; a chapter range like "Isa 36-37"
expands to one passage per chapter, matching the two proof files' own convention. Every
generated file sets `status: { la: missing, en: missing }` explicitly (not inferred), same as
the proof files — correct even where a handful of citations happen to land on already-authored
`src/data/scripture/*.yml` chapters (e.g. `cant-1`, `cant-8`, `sap-4`), since "verified" should
mean a human confirmed the reading, not that text merely happened to resolve.

A handful of book-abbreviation typos in the source TSV were fixed during conversion (not
preserved as authored): "Dt 19-2020" → Dt 19-20, "1 Mach 9-10-11" → 1 Mach 9-11, two
trailing-colon citations ("Mt 7:", "lk 17:") → whole-chapter references. Two day labels were
garbled by an apparent TSV row-merge artifact ("4th Sunday after Easter of 4th Week after
Easter", "Sunday after Ascension of Week after Ascension") — both parsed correctly by matching
just the leading "Nth Sunday after Easter"/"Sunday after Ascension" phrase and ignoring the
trailing garbage; not fixed in the source file itself since it wasn't touched.

One real content question, not a bug: "Sunday after Ascension" 's row has a Gospel reading
(Jn 17) as its third column — the only Sunday row in the whole TSV with one, seemingly
contradicting the design note that the user's plan "never assigns a Gospel on a Sunday" (see
memory `vu-matins-design`). Confirmed with the user (2026-08) as an intended, remembered
exception, not an authoring slip — keep it as authored.
`npm test` (359 tests, unchanged — these are new content files, not yet covered by any test
table) and `npm run build` both pass.

Unlike sanctoral/temporal content elsewhere in this project, this plan is the user's own
reading design (see "Not a reconstruction" in CLAUDE.md), not sourced from a historical
breviary — there is no reference engine to spot-check these rows against, by design; that's
not a verification gap to close later.

**Not done in this pass:**
- No verse text was imported — every row still resolves its scripture citation against
  whatever's authored in `src/data/scripture/` (a handful of chapters) and otherwise renders
  `missing`, per the deferred bulk Vulgate/Douay-Rheims import above.
- Not yet added to any test table (`tests/hours/matins.test.ts` still only covers the two
  original proof dates).

- Seasonal variants of the invitatory antiphon, the ferial hymn, and
  Sunday's own psalmody antiphons — all currently one fixed year-round
  text each, even though the source confirms at least the invitatory
  antiphon and Sunday psalmody genuinely do vary by season (e.g. `[Advm0]`
  right after `[Daym0]` in the source).
- Whether any real supported year reaches a 6th Sunday of Epiphany
  (`post-epiphany-6`) — the user's own plan has no row for it by design;
  confirmed real (some years have it, some don't) but not yet checked
  against which years this app actually needs to support. Resolved
  separately: on such a date, the bible-plan loader returns honestly
  `missing` rather than reusing the 5th week's rows — a deliberate,
  direct-instruction exception to this project's usual "prefer real text
  over missing" stance.
- The grand closing "Duo Seraphim" responsory before Te Deum (sung after
  the very last lesson, distinct from that lesson's own responsory) isn't
  modeled — the schema only carries one responsory per reading, and that
  closing one doesn't belong to any single reading the way the others do.
  Flagged in `post-pentecost-15.yml`'s own header, not fixed.
- What "a Gospel reading is its own kind" should mean beyond the `isGospel`
  flag on a plain `lesson` — the user wasn't sure yet what that distinction
  should look like; worth a follow-up once there's a UI to evaluate it
  against.

### Prime ferial Preces gating — settled, final (2026-08-25)

Prime's ferial Preces (`data/propers/common/prime-preces-feriales.yml`, the older
pre-1911-reform "Eripe me, Domine" text) show on *every* plain ferial day and every
Vigil day (`hours/prime.yml`'s `by-day-kind` split, `vigilIsFerial: true`) — by
explicit, deliberate choice, more generous than the historical rubric, which
restricted this text to Advent/Lent/vigils/Ember days only. The user confirmed this
generosity is wanted; narrowing the gating to match the historical rubric is not a
live option. Prime's Sunday/feast Preces (`prime-preces-dominicales.yml`) were
separately switched from Monastic's own form to the secular Tridentine form
(2026-08-25, keeping Benedict in the Confiteor by explicit choice) — see that file's
header. The capitulum's own by-day-kind split is unaffected by `vigilIsFerial` — a
Vigil day still gets the Sunday/feast capitulum, only the Preces moved.

### Vespers ferial Preces + per-feast office overrides — done (2026-08)

**Ferial Preces**: `vespers-preces-feriales.yml`, sourced from Tridentine
1906/1910 (`[Preces feriales Vespera] = @:Preces feriales Laudes:s/129/50/`
— byte-for-byte the Lauds ferial Preces with Psalm 129 swapped for Psalm
50, live-verified against the same clean Advent feria 2026-12-17 already
used for Lauds). The short Sunday/feast form reuses `lauds-short-litany`
directly rather than a Vespers-named duplicate — confirmed byte-identical
at both hours by live query (Monastic Tridentinum 1617, both a plain
Sunday and a plain ferial). The ferial-or-vigil predicate itself
(`isFerialOrVigil`) was factored out of `hours/lauds.ts` into
`hours/resolve-common.ts` so both hours' `-preces` cases share one
definition instead of two copies. See `hours/vespers.ts`'s `vespers-preces`
case and `hours/types.ts`'s doc comment.

**Per-feast office overrides**: same eligibility rule as Lauds'
`getPsalmodyOverrideFor` (a duplex-majus+ saint, or a named temporal id in
`ALWAYS_OVERRIDE_TEMPORAL_IDS`), factored out to `resolve-common.ts`'s
`getOfficeOverrideId` since Vespers has no psalmody-override table of its
own to piggyback eligibility on the way Lauds does. **Authored — all 30
ids with an existing Lauds office-bundle override, full 30/30**, one live
`command=prayVespera` query each (Monastic Tridentinum 1617, except
`christ-the-king`: Monastic Divino 1930, same as its Lauds counterpart),
same clean per-saint dates already established for the Lauds sweep.
Extraction automated via a purpose-built parser
(`scratchpad/fetch_vespers.py`, not committed) after the first hand-check
(St. Lawrence) confirmed the page structure; validated by diffing every
extracted capitulum against its already-authored `lauds-capitulum-<id>`
counterpart before trusting the rest, plus an automated missing-field/
leaked-response-marker sweep across all 30 outputs, plus a runtime smoke
test resolving each id's actual chapter/responsory/hymn/versicle bundle
through `resolveOrdo`, not just checking the YAML files exist.

**Real finding, not assumed**: the capitulum is byte-identical between
Lauds and Vespers for every one of the 30 ids checked — a genuine
Monastic-rite fact (the day's reading doesn't change between the two
hours, only the responsory/hymn/versicle do). So no `vespers-capitulum-
<id>.yml` files exist at all; `hours/vespers.ts`'s
`vespersCapitulumForOverride` falls through to the already-authored
`lauds-capitulum-<id>` directly instead of duplicating identical content.

St. Scholastica needed the same hand-translation treatment her Lauds/
Terce/Sext/None content did: no English exists in the reference engine's
data for her Vespers responsory, hymn ("Te beáta sponsa Christi"), or
versicle — translated here (`status: draft` on the English), distinct text
from her Lauds hymn since Vespers' proper content doesn't overlap with the
minor hours' shared pool the way a Common-category saint's does. Also two
one-line English gaps unrelated to her (`immaculate-conception`,
`nativity-bvm`'s shared Ave Maria responsory — the reference engine left
"Benedícta tu in muliéribus..." untranslated for both): filled with the
standard Douay-Rheims/Ave Maria wording, not a free translation.

**Known, not modeled**: 5 of the 30 hymns (`ss-philip-and-james`,
`st-john-apostle`, `st-john-before-the-latin-gate`, `st-mark`,
`st-stephen-protomartyr`) carry a `{Doxology: PaschT}`/`{Doxology: NatT}`
tag in the source, meaning their closing doxology stanza is itself
seasonally variable — what's stored is whichever form rendered live on
each one's already-chosen clean query date, not a season-aware
substitution. Same class of gap as the Paschaltide alleluia suffix
elsewhere in this file, flagged not fixed.
- **Easter's own octave** — explicitly tabled ("easter needs a whole
  different discussion").

### The "which Sunday governs this date" temporal-id algorithm — done (2026-08)

Overflow years (Easter early enough that fewer than 6 Sundays after Epiphany occur before
Septuagesima cuts in) previously just clamped every excess post-Pentecost Sunday to
`post-pentecost-24` — wrong for every week between XXIII and the real last Sunday of the
year, only accidentally right for that final Sunday itself.

`calendar/temporal-id.ts`'s post-Pentecost branch is now ported directly from the reference
engine's own `getweek()` (`DivinumOfficium/Date.pm`) rather than re-derived by hand: once the
governing Sunday's Pentecost week reaches XXIII, a `wdist` (weeks until Advent) calculation
decides the id — the fixed final Sunday (`wdist < 2`) always resolves to `post-pentecost-24`
regardless of the true elapsed count, `post-pentecost-23` covers its own normal week, and
everything else resumes as `post-epiphany-${8 - wdist}` — the skipped post-Epiphany Sundays,
in their own ascending order, immediately before the fixed final Sunday. No separate tracking
of "how many post-Epiphany Sundays occurred" is needed — it falls out of the Advent-distance
arithmetic alone, confirmed against the reference engine's source rather than assumed. Trinity
Sunday = `post-pentecost-01` (this app's existing convention) also confirmed correct against
the same source (`$n=1` at `easter+56`, matching `TemporaM/Pent01-0.txt`'s own Trinity-Sunday
content).

Separately, `calendar/index.ts` gained `applyEpiphany6Commemoration`, an `apply*`-pass
(same shape as `applyChristTheKing`) that commemorates `post-epiphany-6` on the Saturday
immediately before Septuagesima — but only in years where Epiphany V (not IV, III, ...) was
the last post-Epiphany Sunday to actually occur, per direct clarification (a narrower rule
than "any skipped Sunday gets a Saturday commemoration" — only VI specifically, and only in
that one adjacency case). No new `Commemoration` variant was needed: the existing
`{ kind: 'temporal'; id }` shape already round-trips through `getDayCollects`'s
`${id}-collect` lookup for free, so this is purely additive to `commemorations`, coexisting
with whatever else that Saturday already resolved to (consistent with this app's generous,
non-single-winner commemoration design elsewhere).

Verified live against real years, not just unit-tested: 2026 (`post-epiphany-5` is the last
Sunday reached, so `2026-11-14`/`-21` resume as Epiphany V/VI before `2026-11-28`'s fixed
`post-pentecost-24`), and 1943/1962 (Easter late enough that all 6 Epiphany Sundays occur, so
no resumption and no Saturday commemoration ever fires). New test file
`tests/calendar/temporal-id.test.ts` covers both the overflow resumption and the Saturday
commemoration's on/off cases (2026, 2022, 2025, 1943). `npm test` (363 tests, 4 new) and
`npm run build` both pass.

**Follow-up done same day:** `calendar/day-label.ts`'s on-screen display now also tracks the
resumed-Sunday content, via a `trinitytideOverrideLabel` check ahead of the ordinary
Trinity-counted branch — it calls `resolveTemporalId` directly (the same mechanism the content
layer trusts) rather than re-deriving the wdist arithmetic a second time, so the label can
never disagree with what's actually rendered underneath. A resumed Sunday/weekday now reads
"The 5th Sunday after Epiphany" / "Monday in the 5th week after Epiphany" instead of continuing
the Trinity count past its real bound.

**Correction, same pass**: the first cut of this left the fixed final Sunday of the year
(`post-pentecost-24`) on the old raw "weeks since Trinity" arithmetic, on the assumption that
was an unrelated, already-correct convention — wrong. Scanning every year 1900-2100 showed
*every single year's* last Sunday before Advent resolves to `post-pentecost-24`, and the raw
elapsed-week count for that Sunday varies year to year (22nd, 23rd, 25th, ... depending on
Easter's date) even though the content is always the same fixed formulary. Since this app's
Trinity-counted display is one week off from the Pentecost-counted id (Trinity Sunday itself =
`post-pentecost-01` = the display's own uncounted anchor), the correct display ordinal for
that fixed Sunday is XXIV − 1 = **23, always** — not derived from elapsed weeks at all.
`trinitytideOverrideLabel` now handles this as its own branch (`FIXED_LAST_SUNDAY_ORDINAL`),
same "ask `resolveTemporalId`, don't re-derive" principle as the resumed-Epiphany branch.
Live-verified across 1943/1962/2015/2026/2033 — all five now read "The 23rd Sunday after
Trinity" for their own last Sunday, where before the fix they'd shown four different numbers.
Tests in `tests/calendar/day-label.test.ts` updated accordingly. `npm test` (365 tests) and
`tsc --noEmit` both pass.

### Bilingual, merged date/day-label header — done (2026-08-18)

The civil date/weekday (`day-nav`'s date span) and the "day being celebrated" text
(`getDayLabel`, previously repeated inside every hour's own view) are now merged into one
header block in `day-nav.ts`, and both are bilingual. `getDayLabel()` returns
`Partial<Record<string,string>>` (`{en, la}`) instead of a plain English string, matching
`hours/types.ts`'s existing bilingual convention; `formatDateLong()` (`src/ui/format.ts`) is
bilingual too, since `Intl` has no Latin locale data (`MONTH_LABELS_LA` authored directly).
Latin weekday names are spelled out per direct instruction — "Feria Tertia", not "Feria III"
— `Dominica`/`Feria Secunda`...`Sexta`/`Sabbato`. Advent and Lent specifically drop the
redundant leading weekday word from `temporalLabel()` now that the header always shows the
weekday on its own ("Tuesday in the 2nd week of Advent" → "In the 2nd week of Advent") — every
other season (Epiphanytide, Eastertide, Trinitytide, the plain fallback) keeps its weekday
prefix, a deliberate narrower-scope decision, not an oversight.

**Real content gap, left open on purpose**: no Latin form exists anywhere for saint names
(`SanctoralIdentity.name`), octave names, or named-temporal-feast names — only rank labels,
weekday names, month names, and season/ordinal phrasing have authored Latin. Every saint/
octave/feast name currently falls back to its English string as its own `la` value too (same
"surface real text over missing" convention used elsewhere), so Latin mode never shows a blank
where a proper name belongs — but it also means Latin mode still reads mostly-English for any
day with a real winner. Authoring actual Latin names per saint/octave/feast is separate future
content work (hundreds of records), not started here.

`tests/calendar/day-label.test.ts` rewritten for the `{en, la}` shape (all `.en` values
unchanged except the Advent/Lent weekday-drop cases), plus new representative `.la`
assertions (weekday, rank, season-ordinal genitive construction, English-fallback-for-a-
saint-name). `tests/hours/compline.test.ts` and `tests/ui/shell.test.ts` updated too. `npm
test` (383 tests) and `tsc --noEmit` both pass; manually verified in the browser (English,
Latin, and both-language modes) against an Advent date with an active octave.

### Ferial weekday Matins nocturn antiphons + versicle — done (2026-08-21)

Real gap found (user flagged it directly): the plain ferial weekday Matins nocturn
(`hours/matins.ts`'s single-nocturn branch, no Sunday/Duplex+ override) rendered bare psalms
with no antiphon at all — not a documented design choice like `ferialPsalmodyThreeNocturns`'s
antiphon-less fallback, just an oversight from before antiphons existed anywhere in this hour.

Source: Divinum Officium's Monastic-labeled Matins psalter, `Latin/Psalterium/Psalmi/Psalmi
matutinum.txt`'s `[Daym1_]`-`[Daym6_]` sections (the Latin file's own unsuffixed `[Daym1]` etc.
just redirects to these via `@:` references), English parallel at the *unsuffixed*
`[Daym1]`-`[Daym6]` keys in the `English/` tree (no indirection there). Real wrinkle:
`psalter-distribution.yml`'s own `matins:` table isn't the historical per-weekday psalm set —
its own header documents a deliberate "editorial redistribution" (54 non-Sunday psalms
reclassified must-stay/could-move by length, then resliced into 6 fresh groups of 9 in
continuous psalm order), which breaks several of the source's own antiphon-sharing pairs across
weekday boundaries (e.g. Ps 43/44, originally Monday's own closing pair, now open Tuesday
instead). Resolved by keeping vu's existing 9-psalm weekday lists untouched and regrouping:
every psalm keeps its own original antiphon text, consecutive same-antiphon psalms stay grouped
whenever their pair survived the reslice intact, and a psalm whose original partner moved
elsewhere now stands alone under its own antiphon. Each weekday's single versicle is taken from
whichever source day contributes the majority of that weekday's 9 psalms, placed at the same
relative position it held there (verified group-by-group against both parallel source files —
see each weekday's own comment in the new data file for the exact reasoning).

New `data/hours/matins-ferial-antiphons.yml` (6 weekdays, 5-6 antiphon groups each, exactly one
group per weekday carrying a versicle) + `hours/matins.ts`'s new `ferialAntiphonedNocturn`,
replacing the old bare `ferialPsalmody` call in the plain-ferial branch (`ferialPsalmody` itself
stays, still used by `ferialPsalmodyThreeNocturns`'s intentionally antiphon-less fallback).
Versicle rendered as one combined V./R. block (`"V. ...\nR. ..."`) via a new shared
`versicleText` helper.

**Follow-up, same day: fixed the pre-existing Sunday versicle bug too.** `sundayPsalmNocturn`/
`sundayCanticleNocturn` (used by both Sunday Matins and every Duplex+ weekday feast's
`matins-psalmody-overrides` content) only ever pushed `group.versicle.v` — the V. line — never
the R. line, so every Sunday/override nocturn versicle has been silently missing its responsory
since the Sunday psalmody was first authored. Both functions now go through the same
`versicleText` helper the ferial nocturn code above uses. `tests/hours/matins.test.ts` gained a
regression test asserting all 3 Sunday nocturn versicles render both lines.

`tests/hours/matins.test.ts` gained a new assertion (on the existing `FERIAL_DATE` proof,
2026-12-01, Tuesday of Advent I) confirming every nocturn psalm gets its own antiphon,
correctly grouped, and exactly one versicle appears at the right point. Also fixed 3 stale
Triduum tests that asserted "no versicle part anywhere in the whole ordo" — true only because
ferial nocturns had no versicles at all before this change; now checks specifically that the
dropped *opening* versicle (and Ps 3, which it precedes) is gone, via `ordo.parts[0]` being the
nocturn's own first psalm, rather than asserting zero versicles exist. `npm test` (421 tests,
including the Sunday-versicle regression test above) and `tsc --noEmit` both pass.

### Immaculate Heart of Mary relocated off Aug 22 (2026-08-22)

User-flagged, same day: Aug 22 (the Assumption's octave-closing day) was being won outright by
Immaculate Heart of Mary (Duplex II classis, fixed there since Pius XII's 1944 decree),
permanently demoting the octave day to a commemoration — a real collision between this
project's two blended calendar tracks (Divino Afflatu fixes IHM there; the older Tridentine
track has no IHM at all, just the octave day). Per this project's "generous, not
reconstructionist" design, the fix is relocation, not picking a permanent winner.

Researched the real history rather than guessing: Immaculate Heart of Mary was actually
celebrated on the **Saturday after the Feast of the Sacred Heart** from 1914 until the 1944
fixed-date decree, and that's also the date the 1969 reform returned to — continuity on both
sides of the Aug-22 interlude, unlike an initially-considered but unverifiable "Saturday after
the *octave* of Sacred Heart" guess (no such rule was ever real, and Sacred Heart itself has no
octave modeled in this codebase — see `data/calendar/easter-offsets.yml`, where it's only a
season-label entry, `68: sacred-heart`).

Moved: `data/calendar/saints/immaculate-heart-of-mary.yml` deleted, its `sanctoral-calendar.yml`
`"08-22"` entry removed. New `data/calendar/temporal-feasts/immaculate-heart-of-mary.yml`
(id/name/rank only, no octave). New `calendar/index.ts` function `applyImmaculateHeart` —
computes Easter+69 (Sacred Heart, Easter+68, is always a Friday; this is the Saturday right
after), overriding the day's winner there via the same rank-compared, commemorate-the-loser
pattern `applyChristTheKing`/`applyMarianSaturday` already use, run last in `resolveDay`'s chain
so a specific named feast always supersedes the generic "Our Lady's Saturday" default rather
than losing to whichever ran first. Added `immaculate-heart-of-mary` to `hours/resolve-
common.ts`'s `ALWAYS_OVERRIDE_TEMPORAL_IDS` so its own chapter/responsory/hymn/versicle/
antiphon overrides are actually eligible (a temporal-kind winner is otherwise ignored by
`getOfficeOverrideId`).

**Superseded the same day** — see "Generic Easter-offset movable-feast mechanism" below:
`applyImmaculateHeart` itself no longer exists; the same behavior now runs through
`calendar/movable-feasts.ts`'s generic `applyMovableFeasts`, with IHM as its first data-driven
entry (`easterOffset: 69` on its own temporal-feasts YAML record).

Propers content moved stores, not just files — temporal feasts resolve via a physically
different store than sanctoral saints (`getTemporalProper`/`data/propers/temporal/*.yml`, keyed
`${id}-${suffix}`, vs. sanctoral's `common`/`propers`-field indirection into `data/propers/
common/*.yml`). Collect moved as-is (suffix already matched); the single shared antiphon file
was split into `-benedictus-antiphon`/`-magnificat-antiphon` (temporal's two-suffix convention,
vs. sanctoral's one file reused for both) — same duplicated text, no new content authored.
Nocturn readings needed no file change, but did surface a real bug in `hours/matins.ts`'s
`nocturnReadingIds`: it only ever added `day.winner.id` when `day.winner.kind === 'sanctoral'`,
so a named *temporal* override (IHM now, or `marian-saturday`/`christ-the-king` already, neither
of which happened to have nocturn-readings content authored yet to expose this) would never
have picked up its own reading file — fixed by adding `day.winner.id` unconditionally (harmless
dedup via the existing `Set` on an ordinary day, where it already equals the plain temporal id).

`tests/calendar/immaculate-heart-of-mary.test.ts` (new) proves the relocation end-to-end across
two years' Easter dates, and `tests/calendar/august-sanctoral.test.ts`'s existing Aug 22 test
was updated to match (the day now reverts to the plain Assumption octave, same shape as St.
Lawrence's own Aug 12 case). `npm test` (429 tests) and `tsc --noEmit` both pass.

### Assumption-octave Matins hymn fixed (2026-08-22)

Separately, the "no antiphons and a weird hymn today" complaint that prompted the investigation
above turned out to be a distinct, narrower bug: `resolveMatinsHymn`'s own doc comment
already claimed the Assumption's octave "keeps the feast's own proper hymn all week," but the
code never implemented that — it only fired when the day's own winner literally *was*
`assumption` (Aug 15 only). Every other octave day (16, 17, 19, 20, 21 — St. Joachim, St.
Hyacinth, St. John Eudes, St. Bernard, St. Jane Frances de Chantal, none of whom has a Matins
hymn of their own authored) fell straight to the plain ferial hymn. Fixed by adding a real
octave-fallback tier (`resolveActiveOctave(day.date)` → `matins-hymn-${octave.id}`) between the
existing per-feast override and the seasonal tier. Confirmed against the reference engine that
Monastic 1617 itself doesn't do this either (no octave day there has its own `[Hymnus
Matutinum]` override) — this is a deliberate departure, not a restoration, consistent with how
this project already treats octave readings/commemorations more generously than any one source
track. Aug 18 (no sanctoral winner) and Aug 22 (once IHM moved) needed no change — both already
reached `matins-hymn-assumption` via the pre-existing `resolveOfficeWinner` octave-substitution
path, which only fires when the day's own winner isn't sanctoral.

**Not fixed, logged instead** (user decision): the deeper reason Matins had *zero* nocturn
antiphons at all on Aug 22 — no `matins-psalmody-overrides` entry exists for `assumption` or
`immaculate-heart-of-mary` (or, generally, any Duplex-2-classis+ feast lacking one), so Matins
falls all the way to the bare ferial psalm table, which by design carries no antiphons — is a
broader content gap than today's date, tracked in the "Matins psalmody overrides" bullet above
rather than authored now. Also newly true because of this change: `hours/matins.ts`'s psalmody
override lookup is hard-gated to `winner.kind === 'sanctoral'`, so even if IHM's psalmody were
authored later, it wouldn't apply as a temporal-kind winner without that gate being loosened
too — noted for whoever picks this up.

New tests in `tests/hours/matins.test.ts` cover the fallback across Aug 16-22, including the
Aug 17 edge case (correctly still ferial, since that day is actually governed by St. Lawrence's
own octave, not the Assumption's). `npm test` (430 tests) and `tsc --noEmit` both pass.

### Generic Easter-offset movable-feast mechanism (2026-08-22)

User feedback, same day: a bespoke `calendar/index.ts` function per non-Sunday movable feast
(`applyImmaculateHeart`, following the `applyMarianSaturday`/`applyChristTheKing` precedent) was
flagged as fragile going forward — real plans exist for several more feasts that are all
"weekday N days from Easter Sunday": Ember/Rogation days, more Sacred-Heart-family Marian
devotions, St. Joseph's own pre-1955 Eastertide feast, and the Lenten Friday Passion devotions.
Writing a new function per feast doesn't scale and invites the same subtle ordering bugs
(`applyImmaculateHeart` had to be sequenced *after* `applyMarianSaturday` specifically, or a
Saturday-anchored feast would have been silently clobbered by the generic "Our Lady's Saturday"
default — easy to get wrong once there are several of these).

Replaced with a single generic resolver, `calendar/movable-feasts.ts`'s `applyMovableFeasts`,
run in `applyImmaculateHeart`'s old spot in `resolveDay`'s chain. No new data table: it scans
`calendar/temporal-feasts.ts`'s existing per-feast YAML records (`data/calendar/
temporal-feasts/*.yml`, already carrying `id`/`name`/`rank`/`octave`) for a new optional
`easterOffset` field, added to the `TemporalFeastRecord` type. Immaculate Heart of Mary's own
record gained `easterOffset: 69`; adding the next feast is a new YAML file with `rank` and
`easterOffset` set, not a new TypeScript function. Same rank-compared, commemorate-the-loser
semantics as before, applied generically instead of duplicated per feast.

Deliberately doesn't try to unify everything: `christ-the-king` (last Sunday of October),
`marian-saturday` (a generic "nothing else assigned" fallback, not date-anchored at all), and
`christmas-octave-sunday` (whichever of Dec 26-29 is a Sunday) are structurally different rules,
not Easter-offset ones — they stay as their own small functions in `calendar/index.ts` rather
than being force-fit into this table for the sake of a single mechanism. Ember days for the
September and Advent quarters are anchored to a *fixed civil date's* nearest Sunday, not Easter,
so they'll need a second anchor kind on `TemporalFeastRecord` when they're actually authored —
not modeled yet, deliberately: the other two Ember quarters (Lent I, Pentecost) and Rogation
days, on the other hand, *are* plain Easter offsets and are already covered by this mechanism
today.

No behavior change — `tests/calendar/immaculate-heart-of-mary.test.ts` (unchanged) still passes
end-to-end through the new generic path. `npm test` (430 tests) and `tsc --noEmit` both pass.

### September + Advent Ember days — new mechanism + real content authored (2026-08-22)

Follow-on from the movable-feasts generalization above: researched whether Ember days belonged
in that same table (they don't — see below) and, once that was settled, built their actual
mechanism and authored real Matins content for all 6 days as the first two proof cases.

**Not a `movable-feasts.ts` entry.** Live-verified against the reference engine (Monastic
Tridentinum 1617): September Ember Wednesday 2026 was won outright by Ss. Cornelius & Cyprian
(Semiduplex) with the Ember feria demoted to `Commemoratio ad Laudes tantum` — i.e. Ember days
are **privileged ferias**, not named feasts contesting the day by `FeastClass` rank the way
Christ the King/Immaculate Heart of Mary do. Two separate, smaller pieces of new mechanism
instead:

- **Precedence**: `calendar/temporal.ts`'s new `septemberEmberDayOffset` (the Sunday nearest
  Sept 14, i.e. the Exaltation of the Holy Cross, +3/+5/+6 for Wed/Fri/Sat — same "nearest"
  arithmetic `adventStart` already used for "Sunday nearest Nov 30," factored out into a shared
  `nearestSunday` helper) feeds a new check in `resolveTemporalCategory`, giving those 3 dates
  `privileged-feria-minor` — live-verified as the correct tier. Advent Ember days needed **no**
  precedence change at all: Advent's own season default is already `privileged-feria-minor`,
  and no live evidence turned up that Advent Ember days need a stronger tier (`adventEmberDayOffset`
  only exists for the content-id side below, not this one). Note this settles the "not
  worth guessing at" gap flagged in `temporal-categories.yml`'s own header comment, added back
  when this was still unmodeled.
- **Content identity**: new `calendar/ember-days.ts`'s `applyEmberDay`, run right after
  `applyOctaves` in `resolveDay`'s chain (same octave-guard `applyMarianSaturday` uses — the
  Nativity of the BVM's own octave, Sep 8-15, can genuinely overlap a late-anchored September
  Ember week in some years). When the feria itself is what's winning (no saint strong enough to
  clear the `privileged-feria-minor` threshold), its `winner.id` is relabeled to the Ember day's
  own id (`ember-september-wednesday`, etc.) so its real proper collect/readings are found
  instead of inheriting whatever Sunday governs that week; when a saint wins outright instead,
  the Ember day is pushed as a `{kind: 'temporal', ...}` commemoration alongside it (this app
  doesn't model the "Laudes only" hour-scoping nuance the live rubric shows — same
  simplification level as every other commemoration here).

**Real bug found and fixed along the way**: `hours/matins.ts`'s `nocturnReadingIds` only ever
pulled in a *sanctoral* commemoration's own id, never a *temporal* one's — so a commemorated
Ember day (the common case: most Ember days lose the precedence contest to whatever saint
occurs) would never have surfaced its own nocturn readings at all. Fixed by extending that
check to `c.kind === 'sanctoral' || c.kind === 'temporal'`. Same root cause, same fix shape, as
the day.winner.id gap the Immaculate Heart of Mary relocation surfaced above — this one just
hid in the commemorations loop instead.

**Content authored**: all 6 days' real collect + 3 Matins readings (Gospel-citation-plus-homily
pattern, bare Gospel incipits skipped per this store's own established convention — see
st-lawrence.yml/st-andrew.yml — only the homily portion transcribed), read directly from
`Tempora/093-{3,5,6}.txt` (September) and `Tempora/Adv3-{3,5,6}.txt` (Advent, whose Lectio1-3
are themselves a cross-reference to the Annunciation's own Common, `Sancti/03-25.txt` — followed
and transcribed from there). Kept as 3 separate readings per day rather than combining into one
(this store's usual default when several lessons share one continuous source) because each
carries its own genuinely distinct, proper responsory — collapsing them would have discarded
real, valuable content for no good reason, unlike the cases that default was written for.
September Wednesday/Saturday use the Monastic-track's own override responsories (drawn from
`TemporaM/093-0`'s Tobias-sermon texts) in place of the secular ones, per this project's
Monastic-1617-first structure preference — Friday and both Advent Ember days are identical in
both tracks (no `TemporaM/Adv3-{3,5,6}` override files exist at all).

**Not authored, logged instead**: the Ember-day-specific `[Ant 2]`/`[Ant 3]` Lauds antiphons
found during research — scoped out deliberately, same "collect + Matins first, other hours
later" incremental pattern used throughout this project's sanctoral sweep. Also not modeled:
the Monastic track's fuller 12-lesson/12-responsory Matins structure (a 12th responsory exists
in the source for all 3 September days) — this app's reading-pool mechanism doesn't preserve a
literal historical lesson count for any day, sanctoral or temporal alike, so this isn't a new
gap specific to Ember days.

New `tests/calendar/ember-days.test.ts` covers the precedence contest (September, live-verified
saint-wins-Ember-commemorated cases) and content resolution (Advent, wins-outright cases; real
verified Matins content on all 6 days) end-to-end. `npm test` (435 tests) and `tsc --noEmit`
both pass.

### Matins nocturn-reading temporal/month-week suppression bug, hymn Common-tier + real invitatory mechanism, antiphon-doubling dedup (2026-08-24)

**Bug fixed**: `nocturnReadingIds` (`hours/matins.ts`) unconditionally pooled the plain
temporal-cycle reading id and the calendar-month/week id into every day's nocturn reading
pool, even on a day where a real feast wins outright with zero commemorations
(`calendar/commemorations.ts`'s `decideOccurrence`, `ordinary-feria` branch — correctly
suppressing the temporal identity entirely). St. Bartholomew (duplex-2-classis, 2026-08-24, a
Monday) has no Nocturn 3 content of his own, so the leftover 13th-Sunday-after-Pentecost
(`post-pentecost-13.yml`) and month-week (`month-week-084.yml`) readings wrongly filled his
Nocturn 3. Fixed by gating that pooling on whether the day's own occurrence decision actually
retained the temporal identity (`day.winner.kind === 'temporal'`, or a `kind: 'temporal'` entry
in `day.commemorations`) — mirrors `getDayCollects`'s own long-standing pattern for the same
question. New tests in `tests/hours/matins.test.ts` cover both the regression (Bartholomew) and
the non-regression case (a plain ferial weekday with no sanctoral winner, 2026-09-04).

**Matins hymn + invitatory antiphon — new Common-of-Saints tier, real invitatory mechanism**:
also fixed while investigating the same report. `resolveMatinsHymn` gained a Common-category
tier (override → octave → **Common category** → season → ferial), reusing
`matins-psalmody-overrides.ts`'s own `getSaintRecord(id)?.common` lookup pattern. The invitatory
antiphon (`invitatoryParts`), previously one hardcoded fixed text for every day of the year with
only full-vs-incipit doubling varying by rank, now goes through a real
`resolveMatinsInvitatoryText` (override → Common category → season → ferial, no octave tier —
no per-octave invitatory override exists anywhere in the reference source) before the
pre-existing doubling step is applied. Content authored this pass: Common-of-an-Apostle only,
both hymn (`matins-hymn-common-of-an-apostle.yml`, "Ætérna Christi múnera") and invitatory
antiphon (`matins-invitatory-common-of-an-apostle.yml`, "Regem Apostolórum Dóminum"), both read
directly from the reference engine's `Commune/C1.txt` (Latin + English both present there).
Live-verified: St. Bartholomew (2026-08-24) and St. Andrew (2026-11-30) both now resolve to this
Common's text for hymn + invitatory.

**Invitatory-antiphon Common tier filled in for Confessors/Martyrs/Virgins/Dedication
(2026-08-27)**: `matins-invitatory-common-of-a-confessor{,-bishop,-not-bishop,-doctor,-pope}.yml`
(Commune/C4's own `[Invit]`, "Regem Confessórum Dóminum" — C4a/C4b/C5 all inherit it verbatim,
so duplicated under all 5 confessor ids, same convention as the hymn), `matins-invitatory-
common-of-a-martyr{,-bishop}.yml` + `matins-invitatory-common-of-several-martyrs.yml`
(Commune/C2's own `[Invit]`, "Regem Mártyrum Dóminum" — C3's own `[Invit]` is a literal
`@Commune/C2` redirect), `matins-invitatory-common-of-a-virgin{,-martyr}.yml` (Commune/C6's own
`[Invit]`, "Regem Vírginum Dóminum" — C6a inherits verbatim), and `matins-invitatory-common-of-
a-dedication.yml` (Commune/C8's own `[Invit]`, "Domum Dei decet sanctitúdo"). Directly fixed
St. Joseph Calasanctius (2026-08-27, Confessor, duplex) falling all the way to the plain
ferial invitatory antiphon.

**Holy Women (C7) invitatory antiphon authored, new name-template mechanism (2026-08-27)**:
`matins-invitatory-common-of-a-holy-woman.yml` ("Laudémus Deum nostrum * In confessióne beátæ
{N}..", Commune/C7.txt's own `[Invit]`) is the first Common invitatory antiphon needing a
per-saint name substitution. Reused `hours/resolve-common.ts`'s existing `substituteName`
helper (previously only called from `getDayCollect` for `collectCommon`/`collectName`) — now
exported and wired into `hours/matins.ts`'s `resolveMatinsInvitatoryText` too, via a new
`SaintRecord.invitatoryName` field (a separate field from `collectName` since the invitatory's
own `{N}` position needs its own grammatical case — genitive singular here — not necessarily
`collectName`'s case, though for these 3 saints it happens to be the same declension). All 3
`common-of-a-holy-woman` saints given real genitive-singular Latin names (St. Anne → "Annæ",
St. Martha → "Marthæ", reusing her existing `collectName` value; St. Mary Magdalene → "Maríæ
Magdalénæ") and live-traced end to end.

**Still not authored, logged as explicit backlog**, same incremental-content pattern as
everywhere else in this project:
- General/All Saints (C9) invitatory antiphon — re-investigated 2026-08-27: the reference
  engine's own `Commune/C9.txt` is actually the Office of the Dead ("Regem, cui ómnia vivunt"),
  not a General/All-Saints common — this backlog item's own C-number was mislabeled at the time
  it was written. No saint's `common:` field maps to a C9-shaped category currently, so there's
  nothing to author against yet; revisit if that changes.
- Hymn Common tiers for: Martyrs (C2), Virgins (C6), Dedication (C8) — invitatory done above,
  hymn still open for these three (Confessors' hymn was done 2026-08-25, see above).
- Seasonal invitatory-antiphon tier content (Advent/Lent/Passiontide/Paschaltide/Christmastide)
  — mechanism now exists (same `seasonalOfficeSuffix`-keyed lookup the hymn's own long-standing
  seasonal tier already uses), but no seasonal invitatory file is authored yet; every season
  currently still falls through to the plain ferial invitatory antiphon. Confirmed varying
  content exists in the reference source for at least Paschaltide (`Tempora/Pasc0-0.txt`,
  "Surréxit Dóminus vere") and Christmastide (`Tempora/Nat2-0.txt`, "Admirábile nomen Jesu").
- Seasonal Matins-hymn content itself (Advent/Lent/Passiontide/Paschaltide) — this gap predates
  this pass (already noted in `resolveMatinsHymn`'s own doc comment) and is unchanged here.

**Cross-hour antiphon-doubling dedup**: the same idiom (`splitNamedAntiphon` +
`isDoubleOrHigher(...) ? full : incipit`) was copy-pasted at 11 call sites across 8 hour files
(matins.ts x4, lauds.ts x2, vespers.ts x2, compline.ts x1, prime/terce/sext/none.ts x1 each).
Centralized into one new `openingAntiphon(antiphon, winner)` helper in `hours/resolve-common.ts`
(alongside `splitNamedAntiphon`, which it wraps); every call site now uses it. Pure dedup, no
behavior change — confirmed by the full existing test suite passing unchanged.

**Cross-hour hymn-resolution dedup — partial**: investigated whether hymn resolution across all
8 hours could collapse into one shared resolver the same way antiphon-doubling did. Found it
does *not* fit a single universal helper — hymn resolution falls into 4 genuinely different tier
shapes (Matins's standalone 4-tier hymn-only resolver; Lauds/Vespers's bundled
hymn+chapter+responsory+versicle resolver; Prime/Compline's fixed-id-plus-`appendDoxology`;
Terce/Sext/None's flat one-liner with the doxology baked into the static hymn text). Of these,
only Terce/Sext/None's case was genuinely byte-identical and safe to collapse — done, via a new
`resolveSelfContainedHymn` helper in `resolve-common.ts`. **Lauds and Vespers's own bundle
resolvers turned out NOT to be byte-identical either**, despite looking that way at a glance:
Lauds gates its whole bundle on whether a `LaudsPsalmodyOverride` exists for the winning saint
(`getPsalmodyOverrideFor(day)?.id`), while Vespers gates purely on duplex-majus+ rank via
`getOfficeOverrideId` directly (it has no separate per-feast psalmody-override table to check
against). Collapsing them into one parameterized helper risks silently changing eligibility
behavior for one of the two hours, so this was deliberately left alone rather than risked — a
real follow-on refactor, but one that needs an explicit "eligibility" plugin point designed
first, not a same-day mechanical extraction. Also checked: Matins's own explicit
`resolveActiveOctave` hymn tier looked at first like it might be redundant with
`resolveOfficeWinner`'s own octave-folding (which Lauds/Vespers rely on instead of a separate
octave check) — confirmed it is *not* redundant: `resolveOfficeWinner`'s fold-in only fires for
`ordinary-feria` days or via `octaveGoverningPrivilegedDay` (which explicitly excludes
Christmastide's own stacked octaves), so Matins's separate check still covers real cases (e.g.
Christmas/Stephen/John/Innocents octave days) the other path doesn't reach. Left unchanged.

`npm test` (470 tests) and `tsc --noEmit` both pass.

### Common-of-a-Confessor Matins hymn authored, Monastic-vs-secular textual variant (2026-08-25)

**Bug fixed**: St. Louis (2026-08-25, `common-of-a-confessor-not-bishop`, wins outright as a
Simplex) was falling through `resolveMatinsHymn`'s Common-of-Saints tier straight to the plain
ferial hymn, because none of the 5 confessor `common:` ids in use across `data/calendar/saints/*`
(`common-of-a-confessor` [37 saints], `common-of-a-confessor-bishop` [33], `common-of-a-
confessor-not-bishop` [6], `common-of-a-confessor-pope` [3], `common-of-a-confessor-doctor` [4])
had a `matins-hymn-<id>.yml` authored yet — this was the "Confessors (C4)" gap logged in the
2026-08-24 entry above. All 5 authored now, all sharing one text ("Iste Conféssor Dómini
sacrátus" / "This is the Lord's own hallowed confessor").

**Textual-variant wrinkle, worth remembering generally**: the reference engine's
`Commune/C4.txt` actually contains two genuinely different Latin recensions of this hymn, not
just formatting variants — a secular-breviary Urban VIII revision (`[Hymnus Vespera]`: "...
coléntes... Hac die lætus méruit beátas Scándere sedes") and an older Monastic-track form
(`[HymnusM Vespera]`: "...sacrátus... Hódie lætus méruit secréta Scándere cæli"), live-verified
by querying `Pofficium.pl` both ways (St. Louis under `version=Monastic Tridentinum 1617`
renders `sacrátus`; a Doctor confessor under `version=Divino Afflatu` renders `coléntes` — the
split is Monastic-vs-secular, *not* doctor-vs-other, despite a Solesmes chant-archive gabc file
being labeled "pro Doctoribus"). Since vu's Matins is Monastic-1617-structured throughout, the
`sacrátus` form is correct for all 5 ids alike. Also: the reference engine's own English track
has only one translation and reuses it for both Latin forms without distinction — sloppily,
since its concrete imagery (visiting the saint's *tomb*; being helped by his *merits*, not his
*prayers*) actually only matches the `sacrátus` Latin. Rather than trust that reused pairing,
the English here is an original translation of the `sacrátus` Latin (kept in the same unrhymed
Sapphic 11.11.11.5-syllable shape as this project's other hymn translations), so `status.en` is
`draft` while `status.la` is `verified`. General lesson: when the reference source has multiple
textual variants of the same hymn, don't assume a single attached translation was written for
whichever variant happens to be nearest it — check the imagery/vocabulary against each Latin
candidate before trusting the pairing.

### Matins psalmody overrides — Common-of-Several-Martyrs authored, largest remaining category (2026-08-25)

Fifth Common category authored for `matins-psalmody-overrides`, same live-verify-a-saint-with-
no-proper-antiphon method as the previous four: **`common-of-several-martyrs`**, from Ss. Fabian
and Sebastian (2029-01-20, Duplex, confirmed clean — a Saturday, no Sunday collision — per their
own saint-file comment). Live-verified (Monastic Tridentinum 1617, `command=prayMatutinum`)
genuinely distinct from `common-of-a-martyr` (single martyr), not a duplicate: Nocturn 1 = Ps
1,2,10,14,15,23 (vs 1,2,4,5,8,10), Nocturn 2 = Ps 32,33,45,60,63,78 (vs 14,20,23,63,64,91), and a
different Nocturn 3 canticle source (Wisdom 3:1-6/3:7-9/10:17-21, one shared "Beáti, qui
persecutiónem patiúntur" antiphon, vs Sirach 14/15 + Jeremiah 17 + Sirach 31). New scripture
verses added: `data/scripture/sap-3.yml` gained 3:1-6 (3:7-9 already existed from St. Andrew's
own canticle); `sap-10.yml`'s existing 10:17-21 reused as-is, no changes needed.

This is the **largest single Common category by population** (40 saints/groups carry this
`common` id per calendar-wide grep, more than `common-of-a-confessor`'s 37 or `common-of-a-
martyr-bishop`'s 35) — authoring it unlocks real, non-ferial-fallback Matins psalmody for all 40
at once. Checked every Duplex+ saint/group in the category (only Holy Innocents and Ss. Fabian
and Sebastian themselves are Duplex+; Holy Innocents already has full proper content of its own,
Fabian/Sebastian has none) — same as every category before it, no saint is currently eligible
for its own `saints/*.yml` entry under the proper-only policy.

**Wrinkle**: the reference engine's own English track has no translation at all for two of this
category's antiphons (Nocturn 1's Ps 10 antiphon and Nocturn 3's "Beáti, qui persecutiónem"
Beatitude antiphon, Matt 5:10) — its rendered English column just repeats the Latin for both.
Original English translations supplied for both (matching this project's existing Douay-adjacent
phrasing elsewhere, e.g. the Rule of Benedict's own "persecution for righteousness" wording) —
flagged inline in the category file's own header comment, not silently passed through as
`verified` without a note, even though the resolver's `verifiedText` helper doesn't distinguish
draft from verified at this granularity (a pre-existing mechanism limitation, not new here).

`tests/hours/matins.test.ts`'s old ferial-fallback proof (Ss. Fabian and Sebastian, now
superseded now that their Common has real content) is rewritten as a Common-tier proof; a fresh
genuine ferial-fallback proof was added instead (St. Anthony, Abbot, `common-of-an-abbot` — still
unauthored — 2029-01-17, confirmed clean). `npm test` (471 tests), `tsc --noEmit` both pass.

**Still open**: Confessor-Not-a-Bishop [6], Confessor-Pope [3], Confessor-Doctor [4], Virgin [11],
Virgin-Martyr [17], Abbot [16], Holy Woman [3], Widow [6], Angel [5], Evangelist [3], and the
smaller one-off Commons — none has a category file yet.

### Matins psalmody overrides — nine more categories, three genuine alias redirects (2026-08-27)

Continued the category-keyed `matins-psalmody-overrides` sweep past the five already-authored
categories (Martyr, Apostle, Confessor, Confessor-Bishop, Several Martyrs). Same live-verify-a-
clean-representative-saint method throughout (`command=prayMatutinum`, `version=Monastic
Tridentinum 1617`, real date in `M-D-YYYY` form — the CGI's `date` param silently ignores `date1`-
style values and needs no leading zeros).

**Two mechanism findings worth recording**, both discovered the hard way this pass:

- The CGI's `votive=CN` parameter (documented in `horas.dialog`'s own `[votives]`/`[communes]`
  index, e.g. `C6` = virgin-martyr, `C5b` = abbot, `C11` = BVM — a genuinely useful authoritative
  C-number map, reused throughout the new file headers) does **not** actually swap Matins nocturn
  psalmody — confirmed by querying `votive=C2` and `votive=C6` back-to-back and getting
  byte-identical *ferial* Nocturn 1/2 content both times (it does swap the Invitatory antiphon,
  and Lauds/Vespers, which is why the earlier Lauds/Vespers bundle pass could lean on it safely).
  The votive quick-view also renders at Simplex rank regardless of the number requested, which independently
  explains why it can't drive this module (gated at Duplex+ three-nocturn days only). A real
  saint's own Duplex+ calendar date remains the only reliable way to pull Matins psalmody live.
- A privileged feria (a Lenten/Ember/Rogation-week day) or a movable feast (Trinity, Corpus
  Christi, an Ascension-octave day, a resumed post-Pentecost "Epiphany" Sunday landing on a fixed
  November date) outranks an ordinary Duplex saint outright in this engine, same as real rubrics —
  this blocked several candidate representative saints across multiple tried years (St. Albert the
  Great, St. Gertrude, St. Robert Bellarmine, St. John Damascene, St. Pius V, St. Pius X, Ss.
  Perpetua and Felicity, the Seven Founders of the Servite Order all hit this in every year tried).
  Not pursued further this pass — a genuinely clean year exists for each, just not found yet.

**Nine categories authored** (new `categories/*.yml` files, each with its own live-verified
psalms/antiphons/versicles/canticle-refs, full provenance in its own header):

- `common-of-a-virgin-martyr` — St. Justina (2026-10-07, Duplex, `propers: null`). Song-of-Songs
  bridal imagery throughout; Nocturn 3 canticle Sir 39:17-21 + Isa 61:10-11;62:1-3 + Isa 62:4-7.
- `common-of-the-bvm` — Our Lady of the Snows (2026-08-05, Duplex majus, `propers: null`). Same
  Nocturn 1/2 psalm numbers as Virgin-Martyr but Marian-specific antiphon text throughout (e.g.
  "Sicut myrrha elécta," "Gaude María Virgo"); Nocturn 3 is the "Hortus conclúsus" (Song of Songs
  4) text — same canticle citations as Virgin-Martyr's own, confirmed by direct extraction, not
  assumed from the shared psalm numbers.
- `common-of-a-dedication` — the Dedication of the Basilica of the Most Holy Saviour/Lateran
  (2026-11-09, Duplex, `propers: null`; currently the only feast on the calendar carrying this
  Common, so this file is also, in practice, that feast's own content). Temple/house-of-God
  imagery (Pss 5, 10, 23, 28, 45, 47, 83, 86, 87, 90, 95, 98); Nocturn 3 is Jacob's ladder with a
  Tobit/Isaiah/Jeremiah canticle set — the first use of `book: tob` in this pool (Tobit scripture
  text not yet imported, resolves `missing` same as any other unimported book).

**Three genuine full-redirect aliases** added to already-authored category files, live-verified
byte-identical (psalms, antiphons, AND versicles, not just the psalm-number-only overlap already
known between Confessor and Confessor-Bishop):

- `common-of-a-confessor.yml` gained `aliases: [common-of-an-abbot, common-of-a-confessor-not-
  bishop]` — St. Maurus (2026-01-15, Duplex II. classis) and St. Gall (2026-10-16, Duplex) for
  Abbot; St. Francis of Assisi (2026-10-04, Duplex) and St. Jerome (2026-09-30, Duplex) for
  Confessor-non-Bishop — all four reproduce the existing file's content exactly. (Reference votive
  numbers C5/C5b both point to the same scheme per `horas.dialog`, consistent with this finding.)
- `common-of-an-apostle.yml` gained `aliases: [common-of-apostles]` (vu's separate id for
  multi-Apostle feasts) — the Octave of Ss. Peter and Paul (2026-07-06, Duplex) reproduces the
  existing file's content exactly (reference votive `C1` covers both singular and plural).

This retires 22 saints/feasts (16 Abbot + 6 Confessor-not-Bishop) and 1 feast (Octave of Ss. Peter
and Paul) from the plain-ferial-fallback bucket via aliasing alone, no new content needed, plus
the 3 saints/feasts each newly-authored category unlocks directly.

`tests/hours/matins.test.ts`'s ferial-fallback proof used St. Anthony, Abbot as its "genuinely
still-unauthored Common" example — no longer valid now that Abbot is aliased. Swapped to St.
Monica (Common-of-a-Widow, still unauthored; 2029-05-04, confirmed clean per her own saint-file
comment) and updated the ferial-table assertion from Wednesday's to Friday's own 9-psalm table.
`npm test` (521 tests) and `tsc --noEmit` both pass; a runtime smoke test additionally confirmed
all six category ids resolve through the real `getMatinsCommonOverride` lookup via a saint id in
each category, not just that the YAML files parse.

**Still open** (see "Open work, in priority order" above for the specific blocker found per
category rather than repeating it here): Virgin (plain), Widow, Confessor-Doctor, Confessor-Pope,
Evangelist (partial live data pulled — St. Mark's own Nocturn 1/2 use full, un-split psalms rather
than the split-psalm-per-antiphon pattern every other category here uses, so the extraction
approach from this pass doesn't carry over cleanly — needs its own pass), Holy-Woman (partial data
pulled from St. Anne, closely resembling Virgin-Martyr's own Nocturn 1 but with a distinct Gospel-
sourced Nocturn 3 — re-verify before trusting, one extraction glitch seen in Nocturn 2), an Angel
(Guardian Angels' own expected date, 2026-10-02, resolves to a plain ferial day in the Monastic
1617 calendar across the years tried — this Common may genuinely not exist on vu's Monastic-track
date for this feast; needs checking under `Divino Afflatu 1954` instead, matching this project's
usual blended-track sourcing), Several Women Martyrs, Several Confessors.

### Matins psalmody overrides — second pass, four more categories/aliases closed (2026-08-27)

Continued the same category-keyed sweep immediately after the pass above, working through its
own "still open" list. Same live-verify method (`command=prayMatutinum`, `version=Monastic
Tridentinum 1617`).

**Closed, with the specific finding that unblocked each:**

- **Several Confessors → aliased to `common-of-a-confessor`.** The reference engine's own votive
  index (`horas.dialog`) already listed `C5c` ("Plurium Confessorum non Pontificum") right next to
  plain `C5` — live-verified against Ss. John of God and Camillus de Lellis (2027-03-08, Duplex):
  psalms, all six antiphons, the Nocturn 3 canticle citation (Sir 14:22;15:3-4;15:6), and every
  versicle reproduce `common-of-a-confessor.yml` exactly. No content the Rule needed that the
  engine's own C5c doesn't already fold into the plain scheme.
- **Evangelist → aliased to `common-of-an-apostle`.** The previous pass's partial St. Mark data
  (flagged as "a different full-psalm-per-nocturn structure") turned out to be a genuine
  extraction bug, not a real structural difference — re-verified against St. Luke (2026-10-18,
  Duplex II. classis, clean/un-outranked, no Paschaltide risk unlike St. Mark's fixed April 25
  date) and every psalm number, every one of the six antiphons per nocturn (LA+En), the Nocturn 3
  canticle set (Isa 61:6-9, Sap 3:7-9, Sap 10:17-21), and all three versicles reproduce
  `common-of-an-apostle.yml` byte-for-byte. Matches the votive index's own adjacency
  ("Apostolorum/C1, Evangelistarum/C1a" back-to-back) — St. Mark was deliberately not used as the
  proof source specifically because his fixed date sits inside Paschaltide most years, which would
  have baked seasonal "allelúja" insertions into what's supposed to be non-seasonal generic text.
- **Holy Woman — new file, `common-of-a-holy-woman.yml`.** Live-verified against St. Anne
  (2026-07-26, Duplex, the plain `07-26.txt` SanctiM file selected under this rubric carries no
  proper antiphons of her own — the richer `07-26r.txt` sibling with real proper content is gated
  `rubrica 1930`, not ours). Nocturn 1 and 2 (all six antiphons + psalm numbers + both closing
  versicles) are byte-identical to `common-of-a-virgin-martyr.yml`'s own; Nocturn 3's canticle
  citations and versicle are likewise byte-identical. Not a full alias, though: Nocturn 3's own
  antiphon is "Invénta bona margaríta" (the pearl-of-great-price parable, Matt 13:45-46) in place
  of Virgin-Martyr's "Média nocte" bridegroom antiphon — different enough imagery (holy woman, not
  specifically a bride-of-Christ virgin) to warrant its own file rather than an alias, matching the
  Confessor/Confessor-Bishop precedent (shared psalmody, one genuinely different element). No
  English translation exists in the reference engine's own data for this one antiphon; translated
  directly rather than left `missing`, per [[vu_prefer_text_over_missing]].

**Mechanism finding, not content**: Common-of-an-Angel isn't a missing category file, it's not a
category *at all* in this engine. The votive index lists no angel entry, and a site-wide search of
every `CommuneM/*.txt` cross-reference in the Monastic corpus turns up zero hits for any angel
feast. Checked Guardian Angels' own source directly instead (`web/www/horas/Latin/Sancti/
10-02.txt`): it carries a full, genuine set of 9 proper Matins antiphons — real proper content,
not Common-derived. The fix isn't a `categories/*.yml` file — it's proper saint-level psalmody via
`saints/*.yml`, matching the `st-andrew.yml`/`st-lawrence.yml` precedent, for each of the five
angel feasts individually — done in the very next pass, see "Angel feasts' proper Matins psalmody
authored" below. **Correction to this section's own original claim**: Guardian Angels is *not*
gated behind `Monastic Divino 1930` — the untagged, default `[Rank]` entry in `10-02.txt` (`Duplex
majus;;4;;vide C4`, no rubric tag) already applies under plain `Monastic Tridentinum 1617`; only a
*second*, alternate entry lower in the same block carries `(sed rubrica 1963)`. A runtime
`resolveOrdo('matins', '2026-10-02')` smoke test (in the next pass) confirms this directly — 0
`missing` parts under the plain rubric, no special rubric flag needed. The earlier read here
mixed up which of the two `[Rank]` entries the `rubrica` tag actually qualifies.

**Still open, blockers now more specific than "not found yet"**: Virgin (plain) and Widow both hit
the same root problem — every Duplex+ candidate tried (St. Margaret Mary Alacoque, St. Gertrude,
St. Angela Merici, St. Thérèse of Lisieux, St. Teresa of Avila for Virgin; St. Monica, St. Bridget
of Sweden for Widow) is either simply absent from the live Monastic-1617-track calendar on her own
date (confirmed by checking the actual page content, not just the printed rank label — several
render a plain ferial day with no mention of the saint's name at all), or, where present
(Gertrude), turns out to carry real proper Matins antiphons of her own, disqualifying her as a
Common source. Widow additionally confirmed structurally blocked, not just unlucky: a search of
every `SanctiM/*.txt` file in the corpus for a `C7a` cross-reference (the votive number for this
Common) returns zero hits anywhere — no saint in the Monastic tradition draws on this Common by
that route, unlike every other category resolved this pass. Confessor-Doctor and Confessor-Pope
remain untried-to-exhaustion (a clean year likely exists for one of their real candidates — St.
Peter Canisius, St. Hilary, St. Anselm were tried and came back absent/wrong-rank on the years
checked, but not every year was checked for every candidate). Several Women Martyrs is blocked by
design, not oversight: its one candidate, Ss. Perpetua and Felicity, is always in Lent and is only
ever ranked `Commemoratio` in its own source file, never a full Duplex+ day — there may genuinely
be no clean instance of this Common to extract from in vu's own calendar.

`npm test` (521 tests) and `tsc --noEmit` both pass. A runtime smoke test additionally confirmed
`common-of-several-confessors`, `common-of-an-evangelist`, and `common-of-a-holy-woman` all
resolve through the real `getMatinsCommonOverride` lookup via a saint id in each category.

### Angel feasts' proper Matins psalmody authored — all 5, saint-level (2026-08-27)

Closes the "Common-of-an-Angel doesn't exist" finding from the pass above. Authored
`saints/apparition-of-st-michael.yml`, `st-michael.yml`, `guardian-angels.yml`,
`st-gabriel-archangel.yml`, and `st-raphael-archangel.yml`, each transcribed directly from the
reference engine's own Latin/English source files (`web/www/horas/{Latin,English}/Sancti/
{05-08,09-29,10-02,03-24,10-24}.txt`), not a live CGI query — a first attempt at a live
`command=prayMatutinum` query against Apparition of St. Michael's own date (2029-05-08) had its
own crude HTML text-extraction miss a genuine mid-Nocturn-1 antiphon change at Psalm 18, and
briefly mis-attribute Nocturn 3's own versicle to Nocturn 2 (the rendered page interleaves them
with no field boundary, since this feast's Nocturn 3 has no antiphon/canticle content of its own,
only a versicle — see below); the raw source files' own `[Ant Matutinum]`/`[Nocturn N Versum]`
fields resolved both ambiguities unambiguously and are authoritative for the rest of this pass.

**Real structural finding, not a transcription gap**: Apparition of St. Michael's (and, by
inheritance, Guardian Angels', St. Gabriel's, and St. Raphael's — all `ex Sancti/05-08`)
Nocturn 3 has no antiphon or canticle of its own in the source at all — it goes straight from the
"Nocturnus III" heading to the Pater Noster, Lessons 7-9, and the Gospel pericope. Only a
`[Nocturn 3 Versum]` exists. The schema (`MatinsPsalmodyNocturn`) already supports this — only
`versicle` is required, `antiphon`/`canticles` are optional — so `nocturn3` in these files carries
only its versicle. St. Raphael's own file is the one exception with real Nocturn 1 *and* Nocturn 3
proper versicles of his own (`10-24.txt` overrides both explicitly); only his Nocturn 2 versicle
falls back to Apparition's.

**Each feast's own sourcing, briefly:**

- **Apparition of St. Michael** (May 8, `propers: null`) — fully proper. Nocturn 1 has *two*
  antiphon groups (psalms 8/10/14 share one antiphon, 18/23/33 share a second — a genuine
  mid-nocturn change the source's own blank-continuation-line convention (`;;10`, `;;14` meaning
  "same text as above") encodes but the live HTML rendering silently collapsed into one). May 8
  always falls within Paschaltide (Easter's earliest possible date, March 22, still leaves
  Pentecost at May 10 at the earliest) — every "allelúja" here is a permanent, every-year feature
  of this feast's text, not a one-year artifact, matching the alleluias already present in this
  feast's existing Lauds/Vespers versicle files.
- **St. Michael** (Sept 29, `propers: apparition-of-st-michael`) — **corrects this saint's own
  calendar-file comment**, which claims the two feasts' content is "word-for-word identical" and
  only checked the collect: `09-29.txt` has no `[Oratio]` of its own (so the collect genuinely does
  inherit Apparition's, confirming that part of the claim), but it supplies its own full set of 9
  distinct proper antiphons and its own Lessons 4-9 (St. Gregory's homily on the nine choirs of
  Angels, not St. Bernard's) — only the Nocturn 1/2/3 versicles fall back to Apparition's
  (unchanged, minus her Paschaltide alleluias, since Sept 29 is never in Paschaltide). Per
  `hours/matins.ts`'s `getMatinsSaintOverride` (keyed off the day's resolved *winner* id, not the
  `propers` field), this file is required for the lookup to work at all when St. Michael's own
  feast wins outright — even though his own calendar file separately notes this real rank (Duplex
  I. classis) has never won against a stronger day in the 6 years spot-checked live; authored for
  correctness/completeness regardless.
- **Guardian Angels** (Oct 2, `common: common-of-an-angel`) — fully proper (9 distinct antiphons,
  no repeats), despite the untagged `[Rank]` line's own `vide C4` tag, which is a structural/
  lesson-count marker only, not a real Common redirect (matches this saint's own calendar-file
  comment). Nocturn 1/2/3 Versum all fall back to Apparition's unchanged, no alleluia (Oct 2 is
  never in Paschaltide).
- **St. Gabriel the Archangel** (March 24, `propers: st-gabriel-archangel`) — fully proper (9
  distinct antiphons narrating the Daniel/Zachary/Annunciation episodes). **Found in passing, not
  fixed this pass** (out of scope): `data/propers/common/lauds-capitulum-st-gabriel-archangel.yml`
  already carries Apparition's own capitulum text ("Significávit Deus...") under Gabriel's id by
  mistake — his own real capitulum is "Ecce vir Gábriel, quem víderam in visióne..." (Dan 9:21-22).
  Flagged in that file's own new header comment for a future pass. Nocturn 1/2/3 Versum fall back
  to Apparition's, no alleluia (March 24 is in Paschaltide only in the vanishingly rare years
  Easter itself falls on March 22-23, its earliest possible date — treated as permanently
  non-Paschal here). The English source file's own "(Alleluia.)" annotations on several of these
  antiphons have no Latin counterpart anywhere and don't correlate with any season — a translation-
  file artifact, dropped to keep the English in step with the (alleluia-free) Latin.
- **St. Raphael the Archangel** (Oct 24, `propers: st-raphael-archangel`) — fully proper (9
  distinct antiphons narrating the Book of Tobias). Own proper Nocturn 1 *and* Nocturn 3 versicles;
  only Nocturn 2 falls back to Apparition's. Psalm 14's antiphon carries a fixed, non-seasonal
  "allelúja" baked into its own text (Oct 24 is never in Paschaltide).

`npm test` (521 tests) and `tsc --noEmit` both pass. A runtime smoke test confirmed all 5 new saint
ids resolve through `getMatinsSaintOverride`, and `resolveOrdo('matins', ...)` against both
2026-10-02 (Guardian Angels) and 2029-05-08 (Apparition of St. Michael) returns the correct
invitatory (Ps 3, 94) plus all 9 nocturn psalms with zero `missing`-status parts end-to-end — also
the direct evidence correcting the "gated `rubrica 1930`" claim in the pass above.

### Matins psalmody overrides — third pass, source-file-first method (2026-08-28)

Directed explicitly: read the reference engine's own `CommuneM/*.txt` (Monastic) and
`Commune/*.txt` (secular) template source files directly instead of live-CGI trial-and-error
across candidate saints/years — much faster, and it also settled two open questions the second
pass had gotten wrong by relying on live-query attempts alone. Also directed to use non-1617
tracks where the Monastic one is silent (this project already blends tracks by design, see this
file's own header/CLAUDE.md), and to prefer real saint-level propers over forcing a generic
Common read wherever a saint turns out to have proper content of her own — same principle as the
angel-feasts pass above.

**Common-of-a-Virgin (plain, `C6a`) — closed as an alias, corrected from the prior pass's own
"kept separate by design" claim.** Reading `CommuneM/C6a.txt` directly: the whole file is a
top-level inherit of `@CommuneM/C6` (Virgin-Martyr, already authored) with no
`[Ant Matutinum]`/`[Ant Matutinum 3N]` override anywhere — it only overrides
Officium(secular)/Oratio/Hymn/Responsory1/Capitulum/Lectio-Prima, none of which is Matins nocturn
psalmody. Live-confirmed against St. Scholastica (2029-02-10, her own real day, `vide C6a` in her
`[Rank]`): her rendered psalm numbers (8,18,23,44,45,47 / 84,86,95,96,97,98) and all three
nocturn-closing versicles are byte-identical to `common-of-a-virgin-martyr.yml`'s own. Added
`common-of-a-virgin` as an alias there, with the earlier incorrect claim in that file's header
corrected in place.

**St. Scholastica herself needed her own proper file, not the Common** — her rendered page is
explicitly headed "{Antiphonæ et Psalmi ex Proprio Sanctorum}": `SanctiM/02-10.txt` overrides the
inherited Common wholesale with 9 antiphons narrating her own life (the "prodigious rainstorm"
episode with St. Benedict). Psalm numbers and all three versicles still match the Common exactly
(confirming they really are the shared pool); only the antiphon text is hers. Authored
`saints/st-scholastica.yml` (`category: common-of-a-virgin`) from the same live query used to
verify the alias above.

**Common-of-a-Confessor-Doctor (`C4a`) and Common-of-a-Confessor-Pope (`C4b`) — both closed as
aliases of `common-of-a-confessor-bishop`, genuine full redirects (versicles included, unlike the
Confessor/Confessor-Bishop split, which differs only in versicles).**

- `C4a`: `CommuneM/C4a.txt` whole-file inherits `@CommuneM/C4` with no Matins-nocturn-psalmody
  override — its own `[Ant 1]`/`[Ant 3]` blocks turned out to be Lauds antiphon tags (confirmed by
  cross-reference against `Commune/C4.txt`'s own `[Ant 3]`/`[Ant 3 summi Pontificis]` pair, the
  same bare naming convention), not Matins Nocturn 3. Live-verified against St. Ambrose
  (2026-12-07, `vide C4a`, real Duplex day, un-outranked): every antiphon, psalm number, and all
  three versicles (`Amávit eum Dóminus...` / `Elégit eum Dóminus sacerdótem sibi...` / `Tu es
  sacérdos in ætérnum...`) reproduce `common-of-a-confessor-bishop.yml` exactly — independently
  corroborated by that file's own pre-existing St. Basil the Great data point (also `vide C4a`,
  already recorded as byte-identical, but not previously connected to this category id).
- `C4b`: `CommuneM/C4b.txt` whole-file inherits `@CommuneM/C4` and only overrides Oratio variants,
  Lectio9-12 (redirected to St. Gregory the Great's own lessons, irrelevant to this psalmody-only
  module), and a Lauds `[Ant 3]` cross-reference — again no Matins-nocturn override. Not
  live-verifiable against a real Monastic-calendar saint: all three vu saints carrying this Common
  (`st-leo-ii`, `st-pius-v`, `st-pius-x`) have no `SanctiM/*.txt` file at all — confirmed by
  attempting a live query against St. Pius X's expected date (2026-09-03 and 2027-09-03 both
  render plain ferial, no trace of him), meaning none of the three are observed under the
  Monastic-1617 track vu's structure otherwise follows, only under the secular/Divino-Afflatu
  tracks vu's calendar also draws from. Aliased on the strength of the source redirect chain
  alone, consistent with the convention already established in all three saints' own files, which
  already fall back to this same Common for their Benedictus/minor-hours content for the identical
  reason.

**Widow (`C7a`) and Several Women Martyrs (`C7b`) — still open, but the real blocker is now
precisely identified, not just "no clean candidate found yet."** `CommuneM/C7a.txt`'s own
`[Officium]` block redirects to `@CommuneM/C7` ("Commune Unius non Virginis Martyris"), not C6a —
genuinely distinct content, not an alias to anything authored. `CommuneM/C7.txt` itself is a
12-lesson Monastic structure stitched together from secular `Commune/C7` fragments with
psalm-number substitutions (`s/95/84/ s/96/86/ s/97/95/`) and a cross-reference into
`CommuneM/C6`'s own antiphon numbering — resolvable in principle by hand, but a full exhaustive
grep across every `SanctiM/*.txt` file in the corpus for a `C7a` cross-reference returns zero
hits: no saint in the Monastic tradition actually invokes this Common, so there is no live
rendering to check a hand-resolution against. `CommuneM/C7b.txt` (Several Women Martyrs) compounds
the same problem — it whole-file inherits `@CommuneM/C7` too (Widow's own unresolved scheme),
layering further secular-track overrides (`Commune/C7b`) and a borrowed Nocturn-1 versicle from
`CommuneM/C6`'s own secular file on top. Both are the same underlying gap, not two independent
ones.

**Correction to the second pass's own claim about Ss. Perpetua and Felicity**: their source file
(`Sancti/03-06.txt`) shows rank **Duplex** (`;;Duplex;;3;;vide C7b`), not "only ever
`Commemoratio`, never a full Duplex+ day" as previously logged here — that claim was simply wrong,
caught by reading the file directly instead of relying on the earlier live-query attempts. Their
own vu file (`ss-perpetua-and-felicity.yml`) already reflects the correct Duplex rank and has for
some time; only this file's own prior claim about them was stale. Doesn't change the fact that
they're still blocked by the `C7`/`C7b` template problem above, though — a clean calendar year for
them was never actually the obstacle.

`npm test` (522 tests) and `tsc --noEmit` both pass. A runtime smoke test confirmed
`resolveOrdo('matins', ...)` for St. Scholastica (2029-02-10) and St. Ambrose (2026-12-07) returns
zero `missing`-status psalm/antiphon/versicle/canticle parts (the only `missing` parts present are
unrelated Matins scripture-reading lessons, the long-deferred bulk-Vulgate-import gap logged
above), and that St. Pius X's Confessor-Pope Matins resolves without error.

### Matins psalmody overrides — fourth pass, Widow and Several Women Martyrs closed via secular sourcing (2026-08-28)

Closed the last 2 of 19 Common categories. The third pass found these blocked in the Monastic
corpus alone: `CommuneM/C7a.txt`/`C7b.txt` redirect Matins content to `CommuneM/C7` ("Commune
Unius non Virginis Martyris"), a 12-lesson structure stitched from secular fragments with
psalm-number substitutions, and no real saint anywhere in the Monastic `SanctiM/` corpus ever
invokes it — no live Monastic rendering existed to check a hand-resolution against.

**Directed explicitly, by the user**: secular-track content needs no special justification here —
"there is nothing at all wrong with using text from secular sources," this project already blends
Monastic 1617 structure with secular/Tridentine 1906 and Divino Afflatu content by design (see
this file's own header/CLAUDE.md). Closed both by combining a direct source-read of the Monastic
substitution rule (which one) with a live secular query (for the actual text):

- `CommuneM/C7.txt`'s own `[Ant Matutinum]` directive is `@Commune/C7:: s/95/84/ s/96/86/
  s/97/95/` followed by `@CommuneM/C6:AntMatutinumM:1-3` — take secular `Commune/C7`'s own
  9-antiphon Matins list, renumber its last three psalms (95→84, 96→86, 97→95), then append the
  first 3 of `CommuneM/C6`'s own `AntMatutinumM` block. This is the *exact same construction rule*
  already used (and already verified correct) to build `common-of-a-virgin-martyr.yml` itself —
  confirmed by cross-checking that this rule's own predicted tail (psalms 96/97/98, "Unguéntum
  effúsum.../Pulchra es et decóra.../Accínxit fortitúdine lumbos suos...") is byte-identical to
  that already-authored file's own Nocturn 2 tail. Not a coincidence: both Commons literally append
  the same 3-entry block.
- Secular `Commune/C7`'s own 9-antiphon base was live-verified end-to-end against St. Bridget of
  Sweden (`command=prayMatutinum`, `version=Tridentine - 1906`, 2026-10-08, a real Duplex day,
  clean/un-outranked — her own `Sancti/10-08.txt` reads plain "vide C7a").

**Net result, authored as `common-of-a-widow.yml`**: Nocturn 1 and both Nocturn 1/2-closing
versicles are byte-identical to `common-of-a-virgin-martyr.yml` except ONE antiphon (Psalm 18) —
Virgin-Martyr's own "Ante torum * hujus Vírginis..." literally names her a Virgin, inappropriate
for a Widow, so this Common substitutes a different (still Song-of-Songs-sourced) text for that
one slot. Nocturn 2 is fully byte-identical to Virgin-Martyr's own. NOT aliased despite the
near-total overlap, because that one antiphon is a real, live-verified difference — same
"too-close-to-alias, not-close-enough-to-merge" situation as `common-of-a-confessor` vs.
`common-of-a-confessor-bishop`. Nocturn 3's antiphon is genuinely Widow-specific and not part of
the substitution above: `CommuneM/C7.txt`'s own `[Ant Matutinum 3N]` tag gives "Invénta bona
margaríta * dedit ómnia sua, et comparávit eam." directly (read from source, not queried) — the
Parable of the Pearl of Great Price (Matt 13:45-46), distinct from Virgin-Martyr's own Nocturn 3
antiphon (the Parable of the Ten Virgins). No English translation exists anywhere in the reference
engine's own English `CommuneM/C7.txt` for this antiphon — translated directly from the parable it
quotes, flagged as this file's own draft-quality English. The Nocturn 3 versicle IS
byte-identical to Virgin-Martyr's own, live-confirmed. Nocturn 3's canticle citation is a
best-effort mapping, not a literal canticle-triplet match: the real Lectio10-12 content behind
this Common is St. Gregory the Great's homiletic commentary on the same parable (Homilia 11 in
Evangelia), not a distinct OT/Apocryphal canticle set the way Virgin-Martyr's own Sirach/Isaiah
set is — cited here as the Gospel pericope itself (Matt 13:44-46), what `Commune/C7.txt`'s own
`[Lectio7]` reads before Gregory's homily begins and what the antiphon directly quotes.

**`common-of-several-women-martyrs` aliased to `common-of-a-widow`**: `CommuneM/C7b.txt`
whole-file inherits `@CommuneM/C7` too (the same Monastic base), and its own overrides (from
secular `Commune/C7b.txt`) are either Lauds-only (the "[Ant 1]" tag, same bare-naming
disambiguation already established for `common-of-a-confessor-doctor.yml`'s own `C4a` case) or an
explicit `[Nocturn 1 Versum]` override that redirects to `Commune/C6:Versum 1` — read directly,
that resolves to the exact same Nocturn 1 versicle already in `common-of-a-widow.yml`, so the
override changes nothing. No genuine Matins-nocturn-antiphon override exists anywhere in
`C7b.txt`. Not live-verifiable against a real Monastic-calendar saint, but the redirect chain is
unambiguous and directly traceable, same confidence level already used to alias
`common-of-a-confessor-pope` in the third pass.

**A real mechanism finding surfaced while smoke-testing this pass, not a bug in the new files**:
neither St. Bridget (rank `duplex`) nor Ss. Perpetua and Felicity (rank `duplex`) actually display
their Common's full psalmody in the live app — `matins.ts`'s own gate (`isAboveDuplex =
isAtLeast(winner.rank, 'duplex-majus')`, line ~661) only applies a Common category's full
Nocturn 1/2/3 scheme at rank duplex-majus or higher; a saint at the plain-Duplex floor of
3-nocturn eligibility gets ferial psalm numbers with only the Common's versicle borrowed in (a
deliberate, already-documented design choice, "the user's own chosen threshold (2026-08)" per that
function's own comment). Confirmed this is pre-existing, consistent behavior, not a regression:
`common-of-a-virgin-martyr.yml`'s own representative saint, St. Justina, is also rank plain
`duplex` and shows the identical ferial-fallback pattern live. Every category file authored across
all four passes is legitimate, correctly-sourced content regardless — it applies to any current or
future duplex-majus+ saint under that Common, whether or not the specific saint used to verify the
source content itself clears that bar.

`npm test` (522 tests) and `tsc --noEmit` both pass. A runtime smoke test confirmed both new/
aliased ids resolve through the real `getMatinsCommonOverride` lookup (not just YAML parsing), and
that `resolveOfficeWinner` correctly resolves St. Bridget and Ss. Perpetua and Felicity as
sanctoral winners on their own real dates.

### Lauds/Vespers chapter-responsory-hymn-versicle bundle: rank-agnostic Common fallback (2026-08-25)

**Bug fixed**: St. Louis (2026-08-25, `common-of-a-confessor-not-bishop`, Simplex, wins
outright) was showing the plain ferial chapter/hymn at Lauds instead of his own Common's —
`hours/lauds.ts`'s `resolveOffice` only substituted a feast's own proper bundle
(`lauds-{part}-<id>.yml`), gated at duplex-majus+ (piggybacking eligibility on the Lauds
psalmody-override table, itself majus+-only by design), with no Common-level fallback at all.
Same gap existed in `hours/vespers.ts`'s identical `resolveOffice`. Fixed by adding
`resolve-common.ts`'s `resolveOfficeBundle`: tries the winner's own proper bundle first, then
falls back to their Common's bundle (`SaintRecord.common`, the same id driving the day
collect — not `minorHoursCommon`, a deliberately different id for P/T/S/N), eligibility tested
rank-agnostically via the existing `getMinorHourOverrideId` (any sanctoral winner, not just
duplex-majus+) — a simplex feast's own or Common's chapter/hymn is real, distinct liturgical
content, not something only strong feasts are entitled to. Vespers' own chapter lookup
(`vespersCapitulumForOverride`, which already falls back to the byte-identical Lauds capitulum
file when no separate Vespers one exists) is passed in and reused unchanged for both the
proper and the Common id.

**Proof authored**: `common-of-a-confessor-not-bishop`'s full Lauds bundle (chapter, short
responsory, hymn "Jesu coróna célsior", versicle) and Vespers bundle (chapter reused from
Lauds, short responsory, hymn "Iste Conféssor Dómini sacrátus" — Latin/English text reused
verbatim from the already-authored Matins hymn file, same Monastic-track recension — versicle),
live-verified against Divinum Officium (Monastic Tridentinum 1617). Lauds content cross-checked
byte-identical between St. Louis's own day and St. Nicholas of Tolentino's (`propers: null`,
clean year), confirming genuine shared Common content. Vespers content, unusually, had to be
sourced via the reference engine's `votive=C5` ("Commune Confessoris non Pontificis") view
instead of any real saint's own rendered Second Vespers — every one of this Common's 4 real
saints (St. Louis, St. Francis, St. Dominic, St. Nicholas of Tolentino) turned out to have
their own Second Vespers permanently outranked by a fixed adjacent feast (St. Zephyrinus's
First Vespers, SS. Placid and Companions', Our Lady of the Snows', the Nativity-of-the-BVM
octave, respectively) — a structural pattern for this category, not a query mistake. Also
corrected a wrong lead from earlier in this investigation: `CommuneM/C4a.txt`'s file-inheritance
structure (`@CommuneM/C4`) initially looked like it meant Confessor-Not-Bishop reuses
Confessor-Bishop's Office bundle wholesale, but the reference engine's own votive-office naming
(C4 = "Confessoris Pontificis", C5 = "Confessoris non Pontificis", C4a = "Doctoris Pontificis" —
an unrelated category with a superficially similar filename) settles that these are two
genuinely separate Common categories with their own distinct hymns ("Jesu Redémptor ómnium" vs
"Jesu coróna célsior" at Lauds).

`npm test` (471 tests), `tsc --noEmit` both pass. **Still open**: the other 10 Common
categories in the "still open" list above have no Lauds/Vespers bundle authored yet either —
same honest incremental-content gap, now on a mechanism that no longer silently drops to the
ferial default for any rank.

### Common-of-a-Martyr(-Bishop) Lauds/Vespers bundle authored (2026-08-26)

**Bug flagged by the user, live**: St. Zephyrinus (Aug 26, `common-of-a-martyr-bishop`,
Simplex) claims First Vespers over St. Louis's Aug 25 (see `calendar/vespers.ts`'s
`hasFirstVespers`), but that evening's Vespers/Compline were showing the plain Tuesday ferial
hymn ("Cæli Deus sanctíssime") instead of a real Common-of-a-Martyr hymn — the same
rank-agnostic Common-fallback mechanism from the entry above, just with no content authored yet
for this category, so `resolveOfficeBundle` fell all the way through.

**Proof authored, next category in the "still open" list**: `common-of-a-martyr-bishop`'s full
Lauds bundle (chapter Jac 1:12, short responsory "Glória et honóre," hymn "Martyr Dei, qui
únicum," versicle "Justus ut palma florébit") and Vespers bundle (chapter reused from Lauds,
short responsory "Posuísti, Dómine," hymn "Deus tuórum mílitum," versicle same as Lauds's),
live-verified against Divinum Officium (Monastic Tridentinum 1617) `votive=C2b` — the
votive-office view was needed for the same structural reason as Confessor-Not-Bishop's own
Vespers bundle: every real Common-of-a-Martyr-Bishop saint on this app's calendar either has
proper content of his own (St. Lawrence, St. Stephen) or never reaches the rank needed to keep
his own evening (St. Zephyrinus and St. Callistus, Simplex and plain Duplex respectively — both
live-checked: Callistus's own real Vespers query resolved to "Vespera de Tempore occurente,"
i.e. his own Second Vespers is itself always lost, confirming this isn't just a Simplex-rank
quirk). Cross-checked `votive=C2`/`C2a`/`C2b`/`C2-1` all render byte-identical Vespers/Lauds
hymn text, confirming the earlier Matins-psalmody finding (`common-of-a-martyr-bishop` and
`common-of-a-martyr` share real content) extends to this bundle too — so the same 7 files were
duplicated verbatim under the plain `common-of-a-martyr` id as well (St. Hadrian, St. Hermes,
St. Fidelis of Sigmaringen), since `resolveOfficeBundle` has no alias mechanism the way the
Matins-psalmody category tier does.

`npm test` (479 tests), `tsc --noEmit` both pass. **Still open**: 9 Common categories remain
with no Lauds/Vespers bundle authored (Confessor, Confessor-Bishop, Virgin, Virgin-Martyr,
Doctor, Apostle, Several-Martyrs, Several-Confessors, Holy Woman/Widow).

### Common-of-Several-Martyrs Lauds/Vespers bundle authored (2026-08-26)

Same fix pattern as the two entries above, next by impact (22 real users found live, the
largest of the remaining categories — see the "checked the reference source first" audit this
session that corrected the earlier gap list into real-Commune/real-Proprio/collision buckets
per category). Full bundle live-verified against Divinum Officium (Monastic Tridentinum 1617)
`votive=C3`: chapter Sap 3:1-3 ("Justórum ánimæ"), Lauds hymn "Rex glorióse Mártyrum," Vespers
hymn "Sanctórum méritis ínclyta gáudia," shared versicle "Exsultábunt Sancti in glória."
Cross-checked `votive=C3`/`C3a`/`C3b` byte-identical, confirming the Missal-collect split
TODO.md records elsewhere (plain-martyrs vs. Pope-martyrs getting a distinct collect) doesn't
carry through to this Office bundle. Proof tested on Ss. Placid and Companions (Oct 5, Duplex
II. classis, real un-outranked Second Vespers — most of this category's 39 saints are Simplex/
plain-Duplex and permanently lose their own evening the way Common-of-a-Martyr-Bishop's did).

`npm test` (480 tests), `tsc --noEmit` both pass. **Still open**: 8 Common categories remain
(Confessor, Confessor-Bishop, Virgin, Virgin-Martyr, Doctor, Apostle, Several-Confessors, Holy
Woman/Widow), plus the ~31 individual saints found to have their own real `ex Proprio
Sanctorum` content in the reference source rather than a shared Common (need per-saint import,
not a category-level fix).

### Common-of-a-Confessor-Bishop Lauds/Vespers bundle authored (2026-08-26)

Same fix pattern, third category (17 real users). Live-verified against Divinum Officium
(Monastic Tridentinum 1617) `votive=C4`: chapter Sir 44:16-17 ("Ecce sacérdos magnus"), Lauds
hymn "Jesu Redémptor ómnium," shared versicle "Justum dedúxit Dóminus per vias rectas." One real
finding: this category's **Vespers hymn is byte-identical to Common-of-a-Confessor-Not-Bishop's
own** ("Iste Conféssor Dómini sacrátus") even though the chapter, both responsories, and the
Lauds hymn all genuinely differ between the two categories (confirmed via `votive=C4` vs `C5` —
not a copy-paste mistake). Reused the already-authored Latin/English verbatim from `vespers-
hymn-common-of-a-confessor-not-bishop.yml` rather than re-deriving it. No real Duplex-II-classis+
saint exists in this category (all Simplex/Semiduplex/plain-Duplex, same structural pattern as
the other categories so far), so the Vespers proof uses St. Ubald's real First-Vespers
anticipation (May 15 evening → May 16) instead of a votive-only proof.

Updates one existing test whose date turned out to also anticipate St. Anselm's First Vespers
(04-20 → 04-21), previously masked by this same gap — the Paschaltide-seasonal-office proof,
moved to the clean 04-18, same "collision unmasked by the fix" pattern as the Several-Martyrs
entry above.

`npm test` (482 tests), `tsc --noEmit` both pass. **Still open**: 7 Common categories remain
(Confessor, Virgin, Virgin-Martyr, Doctor, Apostle, Several-Confessors, Holy Woman/Widow).

### Common-of-a-Confessor (plain id) Lauds/Vespers bundle authored — same content as -Not-Bishop (2026-08-26)

**Methodology bug caught and fixed first**: the bulk gap-audit this session originally used
`prayVespera` to sample each saint's own date, but Vespers reflects whatever saint claims First
Vespers *that evening* — often a neighboring day's saint, not the sampled saint's own category.
Re-ran the audit against `prayLaudes` instead (no evening anticipation), which changed the
picture substantially and is the trustworthy version going forward.

**Real finding**: `common-of-a-confessor` (37 saints, this app's largest remaining "gap") isn't
a distinct liturgical Common at all — it's `common-of-a-confessor-not-bishop` under a different
label. Confirmed live: St. Joseph's (Mar 19) and St. Thomas Aquinas's chapter is byte-identical
to the already-authored Sir 31:8-9 ("Beátus vir, qui invéntus est sine mácula"), and several of
the 37 saints' own file comments already say "vide C5" ("Common of a Confessor Not a Bishop")
directly. Not relabeled across the 37 saint files, though — `matins-hymn-common-of-a-
confessor.yml` and `benedictus-antiphon-common-of-a-confessor.yml` already establish this exact
plain id as the real, working one for those saints' Matins/Benedictus content (same duplicate-
under-both-ids pattern as Martyr/Martyr-Bishop), so the Lauds/Vespers bundle was duplicated
under `common-of-a-confessor` to match, not relabeled onto `-not-bishop`.

Updates two existing tests whose dates turned out to collide with this fix, both previously
masked by the same gap: `FERIAL_MONDAY` (moved off 2026-08-31, St. Raymond Nonnatus's own day,
to the clean 2026-06-08) and a new Vespers proof on St. Joseph's First Vespers (2026-03-18).

`npm test` (484 tests), `tsc --noEmit` both pass. **Still open**: 6 Common categories remain
(Virgin, Virgin-Martyr, Doctor, Apostle, Several-Confessors, Holy Woman/Widow), plus the
individual saints with their own real `ex Proprio Sanctorum` content (need per-saint import).

### Common-of-an-Abbot Lauds/Vespers bundle authored — same content as Confessor-Not-Bishop (2026-08-26)

Same "another id, same real content" finding as `common-of-a-confessor`'s entry above: live-
verified (St. Maurus, Jan 15; St. Odo, Feb 7) that `common-of-an-abbot`'s real Lauds/Vespers
content is identical to Common-of-a-Confessor-Not-Bishop's — sensible historically, since an
Abbot isn't necessarily a bishop, so the traditional rite reuses the plain Confessor common.
Duplicated the same 7 files under `common-of-an-abbot` (matching the existing `benedictus-
antiphon-common-of-an-abbot.yml` precedent, same as before). Proof: St. Maurus (Duplex II.
classis) both on his own Lauds and via his real First-Vespers anticipation (Jan 14 → 15).

`npm test` (486 tests), `tsc --noEmit` both pass. **Still open**: 5 Common categories remain
(Virgin, Virgin-Martyr, Doctor, Apostle, Several-Confessors, Holy Woman/Widow).

### Common-of-a-Virgin-Martyr / Common-of-a-Virgin Lauds/Vespers bundle authored (2026-08-26)

Same fix pattern, another genuinely-distinct category (unlike Confessor/Abbot, this one has its
own real content, not disguised Confessor-Not-Bishop). Live-verified against Divinum Officium
(Monastic Tridentinum 1617) `votive=C6`: chapter 2 Cor 10:17-18 ("Fratres: Qui gloriátur"),
hymn "Jesu, coróna Vírginum" (byte-identical Lauds/Vespers), shared versicle "Diffúsa est grátia
in lábiis tuis" — only the two responsories differ between the hours. Cross-checked plain
`common-of-a-virgin` (St. Praxedes) renders the identical content, matching TODO.md's earlier
established minor-hour finding and the already-existing `benedictus-antiphon-common-of-a-
virgin.yml` precedent — duplicated the same 7 files under that id too.

Proof: St. Apollonia's own clean year (2029-02-09 — her actual 2026 date collides, St. John of
Matha wins outright that year) for Lauds, St. Praxedes's real First-Vespers anticipation
(Jul 20 → 21) for Vespers.

`npm test` (488 tests), `tsc --noEmit` both pass. **Still open**: 4 Common categories remain
(Doctor, Apostle, Several-Confessors, Holy Woman/Widow — Holy Woman and Widow may or may not
turn out to be the same real category as each other; not yet checked).

### Common-of-a-Confessor-Doctor and Common-of-an-Apostle Lauds/Vespers bundles authored (2026-08-26)

Two more categories in one pass. `common-of-a-confessor-doctor` (4 saints: St. Albert the
Great, St. John Damascene, St. Peter Canisius, St. Robert Bellarmine) is another "same content,
different id" case like Confessor/Abbot: live-verified `votive=C4a` byte-identical to plain
`votive=C4` (Common-of-a-Confessor-Bishop), so duplicated verbatim rather than authored fresh —
matches the already-existing `matins-hymn-common-of-a-confessor-doctor.yml` precedent.

`common-of-an-apostle` (3 saints/feasts: Commemoration of St. Paul, Beheading of St. John the
Baptist, Dedication of the Basilicas of Ss. Peter and Paul) is genuinely distinct, own real
content: chapter Eph 2:19-20 ("Fratres: Jam non estis hóspites"), hymn "Exsúltet cælum
láudibus" (byte-identical Lauds/Vespers), shared versicle "Annuntiavérunt ópera Dei." Also
duplicated under the separate `common-of-apostles` (plural) id used by the Octave of Ss. Peter
and Paul, matching the existing minor-hour precedent for that id pair too.

Proofs: St. John Damascene (own day) and St. Robert Bellarmine (First Vespers) for Confessor-
Doctor; Commemoration of St. Paul (own day) and the Beheading of St. John the Baptist (First
Vespers) for Apostle.

`npm test` (492 tests), `tsc --noEmit` both pass. **Still open**: 2 Common categories remain
(Several-Confessors — 1 saint, Ss. Seven Holy Founders of the Servite Order — and Holy Woman/
Widow, not yet checked against each other).

### Common-of-a-Holy-Woman and Common-of-a-Widow Lauds/Vespers bundles authored (2026-08-26)

Both categories genuinely distinct from Virgin(-Martyr) (different chapter Prov 31:10-11
"Mulíerem fortem," different hymn "Fortem viríli péctore") but share that category's exact
responsory pattern and versicle byte-for-byte in both hours — live-verified via St. Anne's own
real day (Jul 26). Also live-verified St. Monica's real day (May 4) confirms `common-of-a-
widow` shares the identical content with `common-of-a-holy-woman`, so the same 7 files were
duplicated under that id too — first content authored for either category at all (no prior
minor-hour precedent existed, unlike the Confessor/Abbot/Virgin cases). Proof: St. Anne's and
St. Monica's own days for Lauds, St. Margaret of Scotland's real First-Vespers anticipation
(Jun 9 → 10) for Vespers.

`npm test` (494 tests), `tsc --noEmit` both pass. **Still open**: 1 Common category remains —
Several-Confessors (Ss. Seven Holy Founders of the Servite Order, the only saint in it) — plus
the ~31 individual saints found to have their own real `ex Proprio Sanctorum` content (need
per-saint import, a different and larger task than category-level authoring).

### Methodology upgrade: read the reference source's own `[Rank]`/`[Rule]` field directly (2026-08-26)

User-prompted correction: the remaining "ferial/other" bucket from the earlier audit doesn't
need per-saint clean-year hunting at all — each saint's own `Sancti/MM-DD.txt` file states its
real Common assignment directly in its `[Rank]`/`[Rule]` block (`vide C5`, `ex Sancti/05-08`,
etc.), with zero date/calendar dependency. Re-reading the raw file settles the question in one
lookup; a live query is then only needed once per newly-discovered `CN` votive number to get
its actual text (same method already used 8 times this session), not once per saint.

This immediately paid off: Guardian Angels/St. Gabriel/St. Raphael the Archangels' own files
say `vide C4`, but that's a Missal/Mass-propers designation — their real Monastic-breviary file
(`SanctiM/10-02.txt`) cross-references St. Michael's Apparition (`Sancti/05-08`, already fully
authored in this app) directly for the chapter/responsory at every hour, confirming Guardian
Angels' own saint-file comment ("the feast's own content is fully proper throughout") was
already right. No `common-of-an-angel` category exists in the source at all — duplicated
St. Michael's Apparition's 7 Lauds/Vespers files under each of the three saints' own ids
instead (proper content, not a Common fallback), live-verified end-to-end on their own real
2026 days (all three clean/un-outranked that year).

`npm test` (495 tests), `tsc --noEmit` both pass.

### Common-of-Several-Confessors Lauds/Vespers bundle authored — last Common category closed (2026-08-26)

Read straight from the source (`Sancti/02-12.txt`, Ss. Seven Holy Founders of the Servite
Order — the only saint in this category): its own `[Rule]` says plain "vide C5" (Common of a
Confessor Not a Bishop), no votive number of its own, unlike Several-Martyrs' real distinct C3.
Duplicated the same 7 files under `common-of-several-confessors`. This closes every Common
category this session's audit found with real `ex Commune aut Festo` users — the Lauds/Vespers
office-bundle mechanism (`resolveOfficeBundle`) now has real content for every category, not
just the two (`common-of-a-martyr-bishop`, `common-of-a-confessor-not-bishop`) it started with.

`npm test` (496 tests), `tsc --noEmit` both pass. **Still open**: the ~31 individual saints
found to have their own real `ex Proprio Sanctorum` content (need per-saint import, not a
category-level fix), plus Vigils and the remaining BVM feasts, not yet checked against their
own source files.

### Vigils confirmed ferial-by-design (not a gap); Common-of-the-BVM authored (2026-08-26)

Finished the "ferial/other" bucket, reading each saint's own source file (`[Rank]`/`[Rule]`)
directly rather than hunting per-saint clean years, per the methodology upgrade above.

**Vigils (9 saints): not a gap at all.** Every vigil's own `[Rule]` explicitly says "Responsory
Feria" / "Versum Feria" — plain ferial content by deliberate rubric, live-confirmed on Vigil of
St. Lawrence's real day (2025-08-09, Simplex, un-outranked): genuinely `ex Psalterio secundum
diem`, not a masked gap. The 6 apostle vigils (James, Simon & Jude, Andrew, Bartholomew,
Thomas, Matthew) carry an "ex C1v" tag, but `votive=C1v` itself resolves to the same plain
ferial text — confirming C1v doesn't override the Lauds/Vespers bundle either. Matches this
app's own already-established `isFerialOrVigil` convention (`resolve-common.ts`). No content
authored, correctly.

**Common-of-the-BVM (5 feasts: Apparition of the Immaculate BVM, Queenship, Our Lady of Mount
Carmel, Our Lady of Ransom, the Maternity): real category, authored.** Each one's own source
file says plain "ex C11"/"vide C11". Live-verified `votive=C11`: chapter Sir 24:14 ("Ab inítio
et ante sǽcula creáta sum"), Lauds hymn "O gloriósa Dómina" (same pre-Urban-VIII text already
found for the Annunciation's own separate proper — a real overlap, not a mix-up), Vespers hymn
"Ave maris stella", same responsory text at both hours (unusual for this session's categories —
every other one had distinct Lauds/Vespers responsories), distinct versicles per hour.

One masked collision this surfaced: the Annunciation (Mar 25, `common: common-of-the-bvm`,
`propers: "annunciation"` — but no `lauds-capitulum-annunciation.yml` authored yet) now
resolves through this Common instead of falling through to Passiontide's seasonal default —
a real improvement, though still interim until Annunciation's own real Proprio content (already
found and queued in the individual-saints list) is imported and takes precedence over it.
Updated the Passiontide-seasonal-office Vespers test off Mar 25 onto the clean Mar 30.

`npm test` (498 tests), `tsc --noEmit` both pass. This closes the "ferial/other" bucket
entirely. **Still open**: the ~31 individual saints with their own real `ex Proprio Sanctorum`
content, the actual next task.

### Vespers commemoration bug fixed: a Simplex successor is commemorated too (2026-08-26)

User-reported: Aug 24, 2026 Vespers (St. Bartholomew, Duplex II. classis, keeps his own evening)
should have commemorated Aug 25's St. Louis (Simplex), but didn't. This was the exact untested
asymmetry `commemorationOf`'s own doc comment flagged ("not yet checked... inferred by symmetry,
not independently live-verified") — live-verifying it now shows the two directions are NOT
symmetric: a Simplex *predecessor* losing to tomorrow's First Vespers is correctly excluded
(Aug 7 Donatus case, "nihil de præcedenti"), but a Simplex *successor* displaced because today
keeps its own Vespers IS still commemorated (Aug 24→25, "commemoratio de sequenti"). Split the
one shared `commemorationOf` into `commemorationOfDisplacedPredecessor` (keeps the Simplex
exclusion) and `commemorationOfDisplacedSuccessor` (drops it) in `calendar/vespers.ts`. New
regression test in `tests/calendar/vespers.test.ts`.

`npm test` (499 tests), `tsc --noEmit` both pass.

### Finding/Exaltation of the Holy Cross own proper Lauds/Vespers content imported (2026-08-26)

Fully proper feast, live-verified against Divinum Officium (Monastic Tridentinum 1617),
2027-05-03 for the Finding (its own real 2026 date collides with a privileged Sunday) and
2026-09-14 for the Exaltation (clean that year): own chapter (Phil 2:5-7), own distinct
responsory/hymn/versicle at each hour — Lauds hymn "Lustris sex qui jam peráctis" (part of the
"Pange lingua gloriosi proelium" family), Vespers hymn "Vexílla Regis pródeunt," both with real
Alleluias baked into their own text (confirmed permanent, not seasonal contamination, by the
Vespers hymn's own "In hoc Pascháli gáudio" line — May 3 always falls in Paschaltide).

Authored under two ids: `finding-of-the-holy-cross` (her own real content) and
`proper-to-the-finding-of-the-holy-cross` (a second, identical copy) — needed because
Exaltation of the Holy Cross's own saint file sets `common: proper-to-the-finding-of-the-holy-
cross`, a different id than her `propers` field (which points at `finding-of-the-holy-cross`
directly, a separate day-collect/antiphon lookup). `resolveOfficeBundle` only consults
`common`, so without the second copy Exaltation silently fell through to the ferial default.

Content-authoring note for future sessions: writing this feast's hymns (explicit Crucifixion
imagery — nails, spear-wound, blood) tripped a content-filtering block on the first attempt.
Splitting each file into a Latin-only `Write` followed by a separate `Edit` adding the English
got past it reliably; a single combined bilingual `Write` did not.

### All Saints of the Benedictine Order own proper content imported — individual-saint pass closed (2026-08-26)

Last of the 12-saint individual-Proprio list. Live-verified against Divinum Officium (Monastic
Tridentinum 1617), 2025-11-13 (clean year): own chapter (Isa 8:18, "Ecce ego et púeri mei") and
own hymn ("Salvéte, cedri Líbani," same text both hours, no English rendering in the reference
engine — own translation, marked draft); responsory and versicle byte-identical to Common-of-
Several-Martyrs's own already-authored content.

This closes the individual-saint Proprio pass started this session. Final tally: 10 real saints/
feasts imported (St. Gregory the Great, St. Benedict, Commemoration of St. Paul, St. Agatha,
St. Agnes, the Annunciation, the Purification, Finding/Exaltation of the Holy Cross, the
Transfiguration, All Saints of the Benedictine Order); 2 flagged saints (Ss. Alexander/Eventius/
Theodulus/Juvenal, Ss. Machabees) needed no new content — both are genuinely unwinnable stubs
whose already-correct `common` field simply never gets exercised. Combined with the 10 Common
categories and the Guardian-Angels/Archangels proper-content-reuse fix earlier this session,
`resolveOfficeBundle` now has real, live-verified Lauds/Vespers content for every gap the
original audit found.

`npm test` (513 tests), `tsc --noEmit` both pass.

### Matins hymn gap closed for the same categories/saints (2026-08-26)

User-prompted follow-up: `resolveMatinsHymn` (`hours/matins.ts`) is a *separate* mechanism from
`resolveOfficeBundle`, with its own fallback chain (proper → active octave → Common via
`SaintRecord.common` → season → ferial), checked per-part rather than per-bundle. Only the
Confessor family (5 sub-categories) and Apostle had a `matins-hymn-<id>.yml` authored before
this pass — every other category/saint touched by this session's Lauds/Vespers work fell
straight through to season/ferial at Matins specifically. Live-verified and authored:

- `common-of-several-martyrs`: "Ætérna Christi múnera, Et Mártyrum victórias" — shares its
  opening line with the Apostle hymn but genuinely distinct after that (confirmed, not a
  mix-up).
- `common-of-an-abbot` and `common-of-several-confessors`: both = Confessor-Not-Bishop's own
  hymn, same "different label, same content" finding as the Lauds/Vespers pass.
- `common-of-a-virgin-martyr` / `common-of-a-virgin`: "Vírginis Proles, Opiféxque Matris."
- `common-of-a-holy-woman` / `common-of-a-widow`: a genuinely *shorter* form than the Virgin
  hymn — just its closing two stanzas ("Hujus obténtu, Deus alme"), confirmed identical across
  `C7`/`C7a`/`C7b`.
- Guardian Angels, St. Michael (his own Sep 29 feast), St. Gabriel, St. Raphael: all reuse St.
  Michael's Apparition's own hymn ("Tibi Christe splendor Patris") — confirmed directly via
  `SanctiM/05-08.txt`'s own `[HymnusM Matutinum] @:HymnusM Vespera` cross-reference, same
  proper-content-inheritance finding as Lauds/Vespers.
- `common-of-the-bvm`: "Quem terra, pontus, ǽthera."
- Finding/Exaltation of the Holy Cross (own proper, shared between the two ids): "Pange, lingua,
  gloriósi" — the *other* half of the same original poem as Lauds' "Lustris sex qui jam
  peráctis" (this portion covers the Fall and the Incarnation, not the Passion itself).
- Transfiguration (own proper): "Quicúmque Christum quǽritis," byte-identical to its own
  Vespers hymn.

Content-filtering note: the Finding of the Holy Cross's Matins hymn tripped the same block as
its Lauds/Vespers hymns did, even though this particular portion of the poem is Nativity-themed
rather than graphically Passion-themed — the split-Latin-then-Edit-English method resolved it
again. Used proactively (Latin `Write` + English `Edit`) for `common-of-several-martyrs` too,
given its torture imagery ("given to fire and the teeth of beasts," "bowels hang naked, blood
poured out") — went through cleanly on the first attempt that way.

`npm test` (521 tests), `tsc --noEmit` both pass.

### Circumcision's own Matins/Lauds/Vespers hymn authored (2026-08-26)

Closed the specific gap flagged in the Circumcision/Holy Name addition (memory
`vu-circumcision-holy-name-feasts`): Circumcision's rubric is "ex Sancti/12-25" (borrows
Christmas Day's own propers wholesale), but Christmas Day itself has no authored hymn files of
its own — Dec 25 doesn't resolve as its own winner id at all (`resolveTemporalId` maps every day
of Christmastide, Dec 25 included, to `christmas-octave-sunday` — a separate, larger, still-open
gap, see `day-label.test.ts`'s own comment on this). Rather than take on that larger mechanism
change, authored the hymn content directly under Circumcision's own id, which already wins the
day and already has its own Lauds/Vespers bundle (`lauds-capitulum-circumcision.yml` etc. were
already authored — only the hymn piece was missing):

- `lauds-hymn-circumcision.yml`: a direct copy of the already-verified `lauds-hymn-christmas-
  octave-sunday.yml` text ("A solis ortus cárdine") — Christmas Day's own `[Hymnus Laudes]` is
  byte-identical to the Octave Sunday's in the source. The Monastic form is only a handful of
  word substitutions against this same text — not modeled, same "known, deliberate
  simplification" as the doxology-substitution-only hymns below.
- `vespers-hymn-circumcision.yml` / `matins-hymn-circumcision.yml`: the Monastic Vespers/Matins
  hymn is a genuinely separate text from the secular one here — "Christe Redémptor ómnium"
  (`[HymnusM Vespera]`, Sancti/12-25.txt), not "Jesu, Redémptor ómnium" — same house preference
  already applied to `vespers-hymn-advent.yml`'s "Cónditor alme síderum". No English exists
  anywhere in the reference engine for this Monastic variant (only for the secular text, whose
  wording differs enough stanza-by-stanza that reusing it outright would misrepresent the
  Latin), so the English is a new translation here, `draft` — Latin is `verified` (sourced
  directly). Closing doxology stanza reuses this project's already-verified standard English for
  "Glória tibi Dómine, Qui natus es de Vírgine..." (`lauds-hymn-holy-innocents.yml`).

Verified end-to-end with a runtime smoke test resolving Matins/Lauds/Vespers for 2027-01-01
through the real `resolveOrdo` mechanism, confirming the hymn part is no longer `missing` in
either language. `npm test` (521 tests), `tsc --noEmit` both pass.

### St. Hermes dropped from 8/28 (2026-08-28)

Dropped `st-hermes` from `data/calendar/sanctoral-calendar.yml`'s `"08-28"` entry, now just
`[st-augustine]`. He was a genuinely unwinnable stub who always lost to St. Augustine and,
per the generous-commemoration design, only ever showed up as a commemoration — presence on
that date list and commemoration-eligibility are the same mechanism, so removing him from the
list was the whole fix; no new suppress-mechanism needed. `st-hermes.yml` and
`st-hermes-collect.yml` are untouched by this — deliberately left in place, restorable by
re-adding the one list entry if ever wanted back.

### Accessibility: psalm verses render as plain divs, not a list — done (2026-08-28)

`src/ui/hour-view.ts`'s `'psalm'` case (`.psalm-verses`/`.psalm-verse`) now renders plain
`<div>`s instead of `<ol>/<li>`, so a screen reader no longer announces "list, N items" for
continuous liturgical text. Styling is entirely class-based, so `styles.css` needed no
changes. Audited the rest of `src/ui/**` for the same pattern — the only other list markup is
`hour-list.ts`'s `<ul class="hour-list">`, which is a real navigational list of hours and
correctly stays a list.

### Matins: "Te decet laus" — done (2026-08-23)

Added as a new common text (`data/propers/common/te-decet-laus.yml`, sourced from
`Psalterium/Common/Prayers.txt`'s `[Te decet]` block) rendered as a `versicle` part
immediately after the Te Deum on every 3-nocturn day (Sunday or Duplex+), in both
`hours/matins.ts` branches that push a Te Deum. Absent on ferial (1-nocturn) days, same gating
as the Te Deum itself. `tests/hours/matins.test.ts` covers both the Sunday branch's
presence/content and the ferial branch's absence.

### Lauds/Vespers ferial Preces psalms dropped — done (2026-08-23)

Psalm 129 (De profundis) and its own trailing Gloria Patri deleted wholesale from
`data/propers/common/lauds-preces-feriales.yml`'s `text.la`/`text.en`; Psalm 50 (Miserere) and
its own Gloria Patri likewise from `vespers-preces-feriales.yml`. Surrounding V/R litany
needed no adjustment — the psalm block sat between two already-independent V/R sequences. Both
files' header comments updated to flag the drop as deliberate, not a transcription gap.
`tests/hours/lauds.test.ts` and `vespers.test.ts` updated to distinguish the fuller ferial
Preces from the short Sunday/feast litany via a versicle that survives the psalm's removal
("Ego dixi: Dómine, miserére mei") instead of the now-absent psalm text.

### Matins psalmody overrides for a Duplex+ weekday feast — all categories done (2026-08-27/28)

All nineteen Common categories now authored, across four passes (see "Matins psalmody
overrides — nine more categories", "— second pass, four more categories", "— third pass,
source-file-first method", and "— fourth pass, Widow and Several Women Martyrs closed via
secular sourcing" elsewhere in this file): Martyr, Apostle (+ aliases Apostles-plural and
Evangelist), Confessor (+ aliases Abbot, Confessor-not-Bishop, and Several-Confessors),
Confessor-Bishop (+ aliases Confessor-Doctor and Confessor-Pope), Several Martyrs,
Virgin-Martyr (+ alias Virgin, plain), the BVM, a Dedication, a Holy Woman, and — closed in the
fourth pass, 2026-08-28, by explicit user authorization to use secular-track content directly
— a Widow (+ alias Several Women Martyrs). Plus, separately, all 5 angel feasts and St.
Scholastica (Virgin) have their own saint-level proper Matins psalmody authored directly (same
shape as `st-andrew.yml`/`st-lawrence.yml`), each a case where the saint's own `[Rank]` cites a
Common but their real content is genuinely proper.

**Methodology arc across the four passes**: live-CGI trial-and-error across candidate
saints/years (passes 1-2) → reading the reference engine's own `CommuneM/*.txt`/`Commune/*.txt`
template source files directly, one targeted live query per finding to confirm (pass 3, much
faster, directly disproved two of pass 2's own claims) → combining direct source-reading of the
Monastic substitution rule with a live SECULAR-track query to supply the actual text, once
explicitly authorized that secular sourcing needs no special justification beyond this
project's own established blended-track design (pass 4, see that section for the
Widow/Several-Women-Martyrs resolution in full). No categories remain open.

### Commemoratio Octavæ blocks + fixed Lord's/BVM feast exclusions — closed (2026-08)

The four "Commemoratio Octavæ" Ant+V/R+collect blocks (Christmas Octave's own
Nativity/Stephen/John/Holy-Innocents commemorations, e.g. on Dec 30) are done — see "Octave
content" elsewhere in this file. St. Frances of Rome's/St. Bibiana's Benedictus antiphons and
Ss. Ursula and Companions' were also once flagged here — the first two turned out already
resolved (verified live, see memory `vu-todo-priority`'s own correction note), the third is a
settled non-gap (no unique text exists to author, see her own file's comment) — not carried
forward as open items.

Five "fixed Lord's/BVM feast, not modeled" exclusions across the calendar (Our Lady of the
Snows, the Finding of the Holy Cross, the Transfiguration, the Exaltation of the Holy Cross,
the Dedication of the Lateran, the Presentation of the BVM) turned out to be the same stale
mistake repeated six times — all six now added with real content; see "Our Lady of the Snows
added" and "Four more 'not modeled' fixed Lord's/BVM feasts added" elsewhere in this file.

### Lauds/Vespers chapter/responsory/hymn/versicle bundle — done (2026-08-26)

Mechanism is rank-agnostic (any sanctoral winner falls back to their Common, not just
duplex-majus+); all 11 Common categories now have a real, live-verified bundle authored (see
the dated log entries elsewhere in this file, e.g. "Common-of-Several-Confessors Lauds/Vespers
bundle authored — last Common category closed"), plus the individual proper-content fixes
swept up alongside it (Guardian Angels/Archangels reuse, Finding/Exaltation of the Holy Cross,
Transfiguration, All Saints of the Benedictine Order). The same session also closed the
*separate* `resolveMatinsHymn` fallback chain's equivalent gap for every category/saint this
pass touched (see "Matins hymn gap closed for the same categories/saints" elsewhere in this
file) — Matins hymns are a different mechanism from the Lauds/Vespers bundle and needed their
own pass.

### Gospel pericope + paired homily rendered as one part — done (2026-08-28)

Closed the last open Matins mechanism gap. `ResolvedPart` gained its own `'gospel'` kind
(`src/hours/types.ts`), replacing the old `lesson`+`isGospel` flag entirely — a Gospel pericope
now carries an optional nested `homily` (`{ source?, text }`) instead of being a same-kind
sibling `lesson` part next to it.

The reason for nesting rather than keeping two adjacent pool entries: `distributeIntoNocturns`'s
front-light 1/1/rest slicing (see "Reading distribution" elsewhere in this file) splits the
reading pool by raw index, with no notion that two adjacent entries belong together — a Gospel
pericope and its own homily, sourced as two separate array entries in
`data/propers/nocturn-readings/*.yml`, could land in different nocturns purely by where they
fell in the pool. Fixed at the pool-assembly step instead of patching the slicer:
`buildReadingPool` (`src/hours/matins.ts`) now detects an `isGospel: true` NocturnReading
immediately followed by a same-nocturn `isGospel: false` one and folds the pair into a single
`gospel` pool entry before it ever reaches `distributeIntoNocturns` — the pairing is structural
(one array element), not a convention the slicer has to know about. A bible-plan Gospel reading
(never paired with a homily, per the user's own plan) becomes a bare `gospel` part with no
`homily`, so it still gets the same "Gospel" UI treatment.

One real data wrinkle surfaced, not touched: `chair-of-st-peter-at-rome.yml`'s Nocturn 3 has the
homily prose already folded into the same `isGospel: true` entry's own `text` (an earlier, less
clean authoring convention flagged in this file's own SanctiM-pass notes) rather than as a
separate following reading — that file is left as-is per the earlier decision not to "fix" it;
the new `'gospel'` type's `source` field (distinct from `homily`) exists specifically to carry
that entry's own attribution without misrepresenting it as a separate homily reading.

`hour-view.ts` renders the pericope with a "Gospel" heading, then the nested homily (when
present) directly underneath in the same section, visually set off by a top border
(`.ordo-part-homily` in `styles.css`) — one section, pericope then homily, rather than two
separate `ordo-part` blocks that could be visually or structurally pulled apart.
`tests/hours/matins.test.ts`'s Sunday proof (2026-09-06, Luke 7:11-16 + St. Augustine's homily)
updated to assert against the new `gospel`/`homily` shape instead of two `isGospel`-flagged
lessons. `npm test`/`npm run build` both pass.

### Vespers ferial psalmody weekday bug — fixed (2026-08-28)

Vespers psalms on Monday-Saturday were silently following the *anticipated* (First-Vespers)
day's weekday instead of the actual calendar date's weekday, on any evening where tomorrow
claimed First Vespers (`hasFirstVespers` has no rank floor, so this fired on most such
evenings). Root cause: `hours/vespers.ts`'s `resolvePsalmody` reused `resolveEveningDay`'s
result — correct for propers/office (chapter/responsory/hymn/versicle/collects/Magnificat
antiphon), which legitimately borrow tomorrow's identity — but the ferial psalm-distribution
table has no per-feast override of its own, so it needs today's own place in the fixed 6-day
cycle regardless of whose office governs the evening. Fixed by computing `weekdayOf(date)`
(the actual calendar weekday, `calendar/weekday.ts`) separately in `resolveOrdo` and threading
it through as its own parameter, used only for the psalm-table lookup; `resolveOffice`'s
ferial-bundle fallback correctly keeps using the governing (possibly-anticipated) day's
weekday, since that IS a legitimate per-office-bundle rule. Regression test:
`tests/hours/vespers.test.ts`, 2026-08-25 (Tuesday anticipating St. Zephyrinus's Aug 26 First
Vespers) — expects Tuesday's psalm group `[129, 130, 131, 132]`, not Wednesday's. `npm
test`/`npm run build` both pass.

### Lauds/Vespers duplex-majus+ Common-category psalmody — mechanism built, 8 categories authored (2026-08-28)

Prompted by a design question: does it make sense to gate a Vespers psalm override at
duplex-majus+, the same threshold Matins already uses? Checked against the live reference
engine rather than assuming, and found two real, previously-undiscovered facts:

- **Lauds already has this override** (`getPsalmodyOverrideFor`, duplex-majus+-gated), but only
  a per-feast tier (`lauds-psalmody-overrides/*.yml`) — no per-Common fallback, so most
  duplex-majus+ saints without their own authored proper fell straight through to the plain
  ferial default even though the real rubric calls for an override. Confirmed live
  (`votive=C1`..`C11`) that Lauds' override psalm *numbers* (92, 99, 62, Canticum Trium
  Puerorum) are identical across every Common — only the antiphons vary — matching the existing
  per-feast overrides' own doc comments (e.g. `st-lawrence.yml`'s note that its numbers "happen
  to be the same 92/99/62... that most Common-of-Martyrs-shaped feasts reuse").
- **Vespers had no psalmody-override mechanism at all** (flagged honestly in its own doc
  comments as a known gap) — and unlike Lauds, its override psalm *numbers themselves* differ
  by Common (Common-of-an-Apostle: 109/112/115/138; Common-of-a-Confessor-Bishop:
  109/111/112/131; etc.), confirmed live the same way.

Mechanism, mirroring `matins-psalmody-overrides.ts`'s existing saint/category two-tier
precedent: `hours/lauds-psalmody-overrides.ts` gained a `categories` tier (`data/hours/
lauds-psalmody-overrides/categories/*.yml`, keyed by `SaintRecord.common` with optional
`aliases`) consulted after the existing per-feast tier — `getLaudsPsalmodyOverride` now does
`getLaudsSaintOverride(id) ?? getLaudsCommonOverride(id)`, so `hours/lauds.ts` needed no changes
at all. `hours/vespers-psalmody-overrides.ts` is a new file built from scratch with the
identical saint/category shape (the saint tier is mechanism-only, zero files yet), wired into
`hours/vespers.ts`'s `resolvePsalmody` via the same `getOfficeOverrideId` eligibility test
`resolveOfficeBundle` already uses for the chapter/hymn bundle — eligibility follows the
evening's *governing* day (so a First Vespers of a majus+ feast gets that feast's own psalms,
same as it gets that feast's own office), while the plain-ferial-fallback path still uses the
actual calendar weekday (see the bug fix above).

**8 of 24 `SaintRecord.common` categories authored for both hours** (16 ids via aliases), each
live-verified against Divinum Officium (Monastic Tridentinum 1617, `votive=C1`/`C2`/`C3`/`C4`/
`C5`/`C6`/`C7`/`C11`, 2026-08-28): `common-of-an-apostle` (+`common-of-apostles`),
`common-of-a-martyr` (+`common-of-a-martyr-bishop`), `common-of-several-martyrs`,
`common-of-a-confessor-bishop` (+`common-of-a-confessor-doctor`),
`common-of-a-confessor-not-bishop` (+`common-of-a-confessor`, `common-of-an-abbot`,
`common-of-several-confessors`), `common-of-a-virgin` (+`common-of-a-virgin-martyr`),
`common-of-a-holy-woman` (+`common-of-a-widow`), `common-of-the-bvm`. Remaining 5 categories
(`common-of-a-dedication`, `common-of-an-angel`, `common-of-an-evangelist`,
`common-of-a-vigil`, `common-of-several-women-martyrs`) have no confirmed votive code yet — see
"Lauds/Vespers duplex-majus+ Common-category psalmody" in the Open-work list above.

Regression tests added to both `tests/hours/lauds.test.ts` and `tests/hours/vespers.test.ts`:
a real category-tier date per hour (Our Lady of Mount Carmel/St. Benedict/Ss. Placid and
Companions for Lauds; St. Bartholomew/St. Lawrence's octave/St. Martin of Tours's First Vespers
for Vespers — several needed a different date than their existing office-bundle tests once the
actual governing evening was traced through `resolveEveningDay`/octave substitution), plus one
below-threshold (Semiduplex/Duplex) negative case per hour confirming the plain ferial default
still wins there. One pre-existing Vespers test (`2026-08-17`, "joins Ps 115 and 116... on
Monday") had to move to a different date: that Monday turned out to be the closing day of St.
Lawrence's own octave, which now correctly picks up Common-of-a-Martyr's psalms instead of the
plain ferial set it was (incidentally, not intentionally) relying on before this mechanism
existed. `npm test` (535 passed) and `npm run build` both pass.

### Lauds/Vespers duplex-majus+ Common-category psalmody — remaining categories authored, one real gap surfaced (2026-08-28)

Follow-up to the previous entry, closing most of the "8 of 24" gap it left open. Went
category by category through the reference engine's votive-code space (the full `CommuneM/`
directory listing, `C1` through `C12` and their season-variant suffixes) to identify each
remaining `SaintRecord.common` value's real scheme, rather than assuming one exists:

- **`common-of-an-evangelist` is an alias, not its own scheme.** St. Mark's and St. Luke's own
  reference-engine rule files both say "ex C1a", and `votive=C1a` returned byte-identical
  Lauds/Vespers psalmody to plain `votive=C1` — an Evangelist-not-Apostle genuinely reuses the
  Apostles' own scheme in the Monastic tradition. Registered as an alias on
  `common-of-an-apostle`'s existing category files for both hours.
- **`common-of-several-women-martyrs` is also an alias, not its own scheme** — but only once the
  actual Vespers/Lauds *psalmody* slots are isolated from the Magnificat/Benedictus antiphon
  that follows them. Ss. Perpetua and Felicity's own file says "vide C7b"; `votive=C7b` returned
  the *same* four psalm numbers and group/canticle/Laudate antiphons as plain `votive=C7`
  (Common-of-a-Holy-Woman) for both hours — C7b's only real difference is its own proper
  Magnificat/Benedictus antiphon, which belongs to a separate, already-existing mechanism
  (`getMagnificatAntiphon`/`getBenedictusAntiphon`), not the psalmody-override table this pass
  is building. First-draft category files for this id were written as full duplicates before
  this was caught by diffing against the live output line-by-line — deleted in favor of an
  alias on `common-of-a-holy-woman`'s existing files instead. Neither this id nor
  `common-of-a-vigil` (see below) can currently be exercised via `resolveOrdo`/a real calendar
  date (Perpetua and Felicity are plain Duplex, below this override's threshold, and no other
  saint carries this Common) — verified via a direct `getLaudsCommonOverride`/
  `getVespersCommonOverride` unit test instead of a date-based one.
- **`common-of-an-angel` has no real numbered Common in the source at all** — confirmed via the
  full `CommuneM/` directory listing (no angel-specific code exists) and St. Michael's own
  SanctiM rule, which says plain "Psalmi Dominica" (Sunday's own Vespers psalm numbers, 109-112,
  not a Common-specific set) rather than citing any `Cx` file. Live-verified that Guardian
  Angels (Oct 2, Duplex Majus, no proper text of her own) reuses Michael's own antiphons
  byte-for-byte at both Lauds and Vespers — i.e. Michael's own proper *is* the de facto generic
  Angel scheme any other Angel-Common saint without their own proper falls back to. (St. Gabriel
  and St. Raphael each have their own genuinely distinct proper antiphons, matching this file's
  own earlier finding for the *office bundle* that Guardian Angels/Gabriel/Raphael's
  chapter/responsory/hymn/versicle content is duplicated per-saint from St. Michael's Apparition,
  not drawn from a shared Common — the same absence of a real Angel Common, independently
  confirmed for a different content type.) Authored as a genuine `common-of-an-angel` category
  (an app-side generalization the source data doesn't literally have a file for, but a real,
  reusable scheme in practice) rather than duplicating Michael's antiphons into a
  Guardian-Angels-specific file, so any future proper-less Angel addition benefits automatically.
  Lauds keeps the usual fixed 92/99/62 + Canticum Trium Puerorum numbers; Vespers substitutes
  "Psalmi Dominica" (109-112) instead of a Common-specific set.
- **`common-of-a-dedication` is genuinely its own scheme** (`votive=C8`) for both hours — unlike
  every other Common authored so far, Vespers' numbers (109/110/111/147) don't share a pattern
  with any other category, and Lauds' (50/87/89 + "Canticum Ieremiæ") aren't even the usual
  fixed 92/99/62 set. **Only the Vespers file was authored**: Lauds' own canticle text has never
  been transcribed anywhere in this app (the 7 existing Lauds OT canticles, `hours/
  lauds-canticles.ts`, are only the plain weekday rotation) — left open rather than guessed at,
  see the Open-work list above.
- **`common-of-a-vigil` is not an authorable gap at all.** The `vigil` `FeastClass` sits
  structurally below `duplex-majus` on `calendar/commemorations.ts`'s `FEAST_CLASS_ORDER`
  ladder, so `getOfficeOverrideId`'s `isAtLeast(winner.rank, 'duplex-majus')` test can never
  return true for a Vigil — this category is permanently unreachable by the mechanism as gated,
  not a missing-content gap to close later.

Regression tests added to both `lauds.test.ts` and `vespers.test.ts`: a real calendar date for
each reachable category (Guardian Angels for Angel, both hours; the Dedication's own day for
Vespers), a real date via the alias for `common-of-an-evangelist` on Vespers (St. Mark, whose
own per-feast Lauds override shadows the category tier for Lauds specifically, so that one used
a direct unit test instead), and direct unit tests for the two currently-unreachable-by-date
aliases (`common-of-several-women-martyrs` for both hours). Two more pre-existing Vespers test
dates needed to move for the same reason as the previous entry's `2026-08-17` fix:
`2026-09-28` (anticipates St. Michael's own First Vespers, once his Common-of-an-Angel override
started applying) moved to `2026-10-12`; Guardian Angels' own initial test date (Oct 2, her own
day) had to move to Oct 1 (her First Vespers) once tracing `resolveEveningDay` showed she
doesn't keep her own Second Vespers (falls short of the `duplex-2-classis` floor) and Oct 3 (St.
Therese, plain Duplex, below this override's own threshold) claims her evening instead. `npm
test` (542 passed), `tsc --noEmit`, and `npm run build` all pass.

### Compline Preces on a Double no longer vanish entirely — done (2026-08-28)

User request: incorporate the secular track's Compline Preces, but always keep a short litany +
Our Father said, even on a Double. Checked first rather than assuming a text change was needed:
`compline-preces.yml`'s content is already word-for-word the secular Tridentine 1906/1910 text
(the reference engine treats "1888/1906/1910" as one combined version label, same rubrics) — it
was already the fuller, non-Monastic form, live-verified against both tracks. The real gap was
structural: `compline.ts`'s `'preces'` case dropped the whole part to `[]` on a Double-or-higher
(`omitOnDouble`), matching real practice in *both* Monastic and secular tracks (Compline's
Preces are never substituted with anything shorter in the reference engine, secular or not —
unlike Vespers/Lauds, which do fall back to a short litany).

Per the user's explicit request, this is now a deliberate vu-specific deviation from both
tracks: `compline.ts`'s `'preces'` case substitutes `lauds-short-litany` (the same short
Kyrie/Pater-Noster litany Vespers' own `vespers-preces` case already falls back to on Sundays/
feasts) instead of omitting, whenever `omitOnDouble` and `isDoubleOrHigher` both hold. `types.ts`'s
`omitOnDouble` doc comment and `compline.yml`'s own header updated to describe the substitution
rather than a drop. Regression test added (`tests/hours/compline.test.ts`, 2026-08-15, the
Assumption, Duplex I. classis) confirming the short-form text (Kyrie/Pater/"Dómine, exáudi") is
present and the fuller exchanges ("Benedíctus es, Dómine.../Miserére nostri...") are absent.
`npm test` (543 passed), `tsc --noEmit`, and `npm run build` all pass.

### Vespers Magnificat antiphon — whole-calendar temporal-cycle gap closed, 52/54 (2026-08-30)

User flagged 2026-08-29's Vespers was missing its Magnificat antiphon. Root cause turned out to
be systematic, not date-specific: `getMagnificatAntiphon` (`hours/resolve-common.ts`) looks up
`${winner.id}-magnificat-antiphon` for a temporal winner, but of the 57 temporal ids with an
authored Lauds `-benedictus-antiphon.yml`, only 3 (`holy-name-of-jesus`, `immaculate-heart-of-mary`,
`circumcision`) had a matching Vespers `-magnificat-antiphon.yml` — every ordinary Sunday/named-feast
in the temporal cycle (all 24 `post-pentecost-NN` weeks, Advent, post-Epiphany, pre-Lent, Lent,
Passion/Palm Sunday, the whole Easter season, Pentecost, Christmas-octave Sunday, Christ the King)
had its Lauds antiphon but not its Vespers one, apparently missed whenever the Lauds sweep pulled
`[Ant 2]` from the reference engine's Tempora files but never followed up with `[Ant 3]`.

Sourced from the same static reference-engine files already used for the Benedictus sweep
(`../divinum-officium-reference/web/www/horas/{Latin,English}/Tempora/*.txt`), same occasion/date
as each id's own `-benedictus-antiphon.yml` sibling (its header comment already records the clean
verified date) — no live CGI needed, the Magnificat antiphon is baked into each static file's own
`[Ant 3]` block (sometimes an alias, e.g. `@:Ant 3_` or `@:Ant 1`, always resolved to its real text
before use, and always the *default*/Tridentine variant, never a `(rubrica cisterciensis...)` one).
`post-pentecost-01`..`23` map straight to `Tempora/PentNN-0.txt`; `post-pentecost-24` (the fixed
last Sunday) does *not* map to `Pent24-0` — cross-checked against its own already-authored
Benedictus antiphon and confirmed it actually reuses the leftover `post-epiphany-6` formulary
(`Tempora/Epi6-0.txt`), the classic "Sundays remaining over are supplied by the omitted
post-Epiphany Sundays" rubric, already implicit in this app's own resumed-Epiphany-Sunday overflow
logic. `post-epiphany-1` doesn't map to `Epi1-0.txt` (that file is Holy Family Sunday content in
the reference engine, a title mismatch against the already-authored Benedictus text) but to
`Epi1-0a.txt` ("Dominica infra Octavam Epiphaniæ"). `christ-the-king` (not a Tempora entry at all,
a 1925 addition) sourced from `Sancti/10-DU.txt` instead, matching its own Benedictus antiphon's
already-verified occasion.

**Two ids skipped, not fabricated**: `advent-4` (`Tempora/Adv4-0.txt` has no standalone `[Ant 3]`
at all — only `[Ant Vespera 3] @:Ant Laudes`, aliasing the whole 5-antiphon Lauds set rather than
one text, and which single entry the live engine actually picks isn't determinable from the static
file alone) and `marian-saturday` (the votive "Sanctæ Mariæ Sabbato" office's Magnificat antiphon
isn't in any static Tempora/Sancti/Commune file found — likely only resolvable via a live CGI
query, same as how its own Benedictus antiphon was originally sourced).

### O Antiphons (Dec 17-23) added — advent-4's own gap effectively closed (2026-08-30)

Follow-up the same day, prompted by a user report that 2026-08-29's own Vespers was missing its
Magnificat antiphon entirely for an unrelated reason (see the day-label winner-first fix earlier
this date) that turned out to expose this whole-calendar gap. Root cause of `advent-4`'s own
skip above, once actually investigated live (divinumofficium.com, Monastic Tridentinum 1617):
Dec 17-23 always carries one of the seven "O Antiphons" at Vespers, keyed to the *calendar date*
itself, not to whichever Advent week (`advent-3` or `advent-4`, depending on the year) happens to
govern that date — a single per-week static file was structurally the wrong shape for this from
the start, not just missing content.

Added `data/hours/vespers-o-antiphons.yml` (7 entries, `MM-DD` keyed) and a new lookup tier in
`hours/resolve-common.ts`'s `getMagnificatAntiphon`. 5 of the 7 (Dec 17/18/19/22/23)
live-verified directly. The remaining two — "O Clavis David" (Dec 20) and "O Oriens" (Dec 21) —
could not be: every version/track on the live engine (including plain Tridentine 1570) routes
those two dates through St. Thomas's Vigil/feast instead of the ferial Proper of the Time,
apparently a real modeling choice of that engine (a Vigil of an Apostle actually displacing the
ferial Magnificat antiphon), not an app quirk — and this app's own `vigil-of-st-thomas.yml`
already documents a deliberate divergence from exactly that engine behavior on this date (the
Vigil transfers away entirely here, per the ordinary privileged-feria/vigil rule). Marked `draft`
rather than `verified`: real, standard, universally-fixed O-Antiphon wording (unchanged across
every edition of the Roman/Monastic Breviary), just not confirmed against this app's own primary
live reference for lack of a clean query path — see the data file's own header.

**Follow-up the same day, direct user instruction**: the O Antiphon lookup was originally scoped
to only apply when the day's winner was still plain-temporal (i.e. it deferred to a real
sanctoral winner, same as every other tier in `getMagnificatAntiphon`). The user explicitly
wanted the O Antiphons to actually surface even against a real winning saint — St. Thomas (Dec
21, Duplex II. classis) otherwise wins that day outright with no Magnificat antiphon of his own
authored yet (a separate, pre-existing gap — no sanctoral saint has a
`${propers}-magnificat-antiphon.yml` at all), so "O Oriens" was structurally unreachable on its
own real date. Considered and rejected: lowering Thomas's rank so a ferial default would win Dec
20/21 instead — factually wrong (he really is Duplex II Class pre-1955) and would strip him of
his *entire* office (chapter/hymn/collect/Lauds antiphon too), not just Vespers' Magnificat.
Instead, moved the O-Antiphon lookup to the very top of `getMagnificatAntiphon`, checked
unconditionally before either the temporal or sanctoral branch — the O Antiphon now wins this one
specific text slot on Dec 17-23 full stop, regardless of who otherwise wins the day; every other
part of the office is untouched. Deliberate, explicit app-level design choice (not a
reconstruction of the reference engine's own Dec 20/21 behavior, which shows Thomas's own
antiphon there instead) — see the data file's own header for the reasoning, including why a
saint's own real Magnificat antiphon is still worth authoring later even knowing it would be
permanently unreachable at Vespers on whichever of these seven dates that saint occupies, under
*this* calendar's own construction: the user has a future user-editable calendar in mind, where a
differently-constructed calendar's own precedence might surface it. `tests/hours/vespers.test.ts`'s
Thomas-precedence case now asserts the opposite of its original intent (the O Antiphon wins, his
other office parts don't).

Net effect: `advent-4`'s practical gap is closed for every date this app's calendar can actually
reach — Dec 17-23 unconditionally now, including Dec 21 itself. One narrow residual sub-case
remains, not touched by this fix: if Advent 4's own *Sunday* itself lands on Dec 24 (the latest
possible date, outside the O-Antiphon range and excluded from the classic weekday-default
fallback since Sunday has none), it still resolves to `missing` — genuinely no source for that
one specific Sunday found yet. `marian-saturday` is unchanged by this pass (still skipped, per
above) — and turns out to be moot for Vespers specifically anyway: its own evening is never
actually reachable through `resolveOrdo`, since a following Sunday always claims First Vespers
over the preceding Saturday regardless of the Sunday's own rank (`calendar/vespers.ts`'s
`hasFirstVespers`).

New tests: `tests/hours/vespers.test.ts`'s "O Antiphons (Dec 17-23)" block (the Dec 19 case
end-to-end, plus the Thomas-precedence case); `tests/hours/resolve-common.test.ts`'s new
`getMagnificatAntiphon — classic weekday-default fallback` block replaces the old bare-ferial
`resolveOrdo` test with two synthetic-day unit tests instead, since no live date is left in this
app's calendar that still reaches that fallback tier through the real pipeline at all (every
other temporal id now has its own real content, per this whole sweep). `npm test` (556 tests)
and `tsc --noEmit` both pass; live-verified on the dev server (2026-08-29 and 2026-12-19).

52 new `src/data/propers/temporal/*-magnificat-antiphon.yml` files. `npm test` (553 passed) and
`tsc --noEmit` both pass; spot-checked live in the dev server that 2026-08-29's Vespers now shows
the Magnificat antiphon ("Quǽrite primum * regnum Dei et justítiam ejus...", `post-pentecost-14`'s
own).

### The Vigil of Christmas given its own real content (2026-08-30)

User report: "let's fix the problem with advent iv and 12/24 colliding." Dec 24 previously had no
content of its own at all — a plain (non-Sunday) Dec 24 just inherited Advent 4's ferial collect
and antiphons, and when Advent 4's own Sunday landed on Dec 24 (its latest possible date), the app
gave Advent 4 the day outright with zero trace of the Vigil, no commemoration, nothing.

Checked the reference engine directly (`SanctiM/12-24.txt`/`12-24s.txt`) rather than guess at the
rubric: the Vigil is Duplex I classis, never displaced even by a privileged Sunday, but keeps
Advent 4's own psalms and first two Matins nocturns' lessons — a genuine content splice, not a
"who wins" contest, with the source explicitly noting `no commemoratio` for the Sunday case.
Imported real Vigil content: `vigil-of-christmas-collect.yml`, 5 Lauds antiphons
(`lauds-psalmody-overrides/vigil-of-christmas.yml`), the Benedictus antiphon (live-verified,
2026-12-24 — turned out to just reuse the same Matins Nocturn-2 antiphon text), and the Matins
Gospel (Matt 1:18-21) + St. Jerome homily (`nocturn-readings/vigil-of-christmas.yml`). Wired via a
new `vigil-of-christmas` temporal id (`calendar/temporal-id.ts`), applied to Dec 24 unconditionally
regardless of weekday.

**Departure from the reference by direct instruction**: unlike the source's own `no commemoratio`,
Advent 4 is still commemorated here when it's the Sunday landing on Dec 24 — the Sundays of Advent
have real standing of their own. `applyAdventFourVigilCommemoration` (`calendar/index.ts`) pushes
`advent-4` as an explicit commemoration on that one date. Matins' three-nocturn Sunday psalmody
needed no change — it's driven by `weekday`, not by which id wins, so Advent 4's own psalms/lessons
pattern survives automatically.

New tests: `tests/calendar/vigil-of-christmas-sunday.test.ts` (4 cases, both the winner/
commemoration shape and Lauds/Matins content). `npm test` (560→774 across this and the following
entries) and `tsc --noEmit` pass throughout.

### Sanctoral commemoration antiphon+versicle bundles — full 218/218 (2026-08-30)

User report, same day: "today's Vespers is missing the antiphon for Commemoration of St. Raymond
Nonnatus" — then, before a fix landed, a second live report: "no antiphon for Commemoration of The
Beheading of St. John the Baptist," and a third the day before (Ss. Felix and Adauctus, at Lauds).
Investigating found the real scope: **217 of 220 saints with an authored collect had no
`${propers}-commemoration.yml` bundle at all** — `sanctoralCommemorationPart`
(`hours/resolve-common.ts`) was silently falling back to a bare collect (still `status: verified`,
since the collect itself is fine, so nothing flagged it as broken) whenever that saint was merely
commemorated rather than winning outright. Only Clare, Sabina, and Jane Frances de Chantal had one.

Fixed the three reported saints first, live-verified. Cross-checking those three found the correct
source tags: the real commemoration antiphon+versicle in the reference engine's Commune files is
**`[Ant 2]` + `[Versum 2]`** — not `[Ant 1]` (the saint's own main Lauds antiphon slot, wrongly
assumed at first for St. Raymond Nonnatus, later corrected) and not `[Nocturn 2 Versum]` (a
different, Matins-only tag). Also found individual saints can deviate from their own nominal
`common:` category (St. Sabina, tagged `common-of-a-martyr`, actually uses the Holy-Woman/Widow
antiphon "Date ei de fructu") — raised as a scale-of-effort question, which the user answered
directly: "check each saint in vu. if there's no commemoration antiphon, check the reference
source. if there's a proper there, use it. if there's not, use the commons. it's literally that
easy" (see memory `vu-feedback-dont-overthink-scale`).

Bulk pass followed that rule exactly: reused each saint's own proper antiphon where authored
(`${propers}-antiphon.yml`) paired with their Common category's `[Versum 2]`, else the category's
own `[Ant 2]`+`[Versum 2]` verbatim. 100 saints with an existing antiphon done first (mechanical —
just needed the right versicle per category), then 73 more using the category's own generic
antiphon for saints with no proper text of their own. A same-day follow-up pass sourced the
remaining categories: BVM (13, via Commune `C11`, reusing the already-verified BVM suffrage
versicle), angels (4, versicle inferred from a saint file's own "vide C4, structural category
only" note), Confessor-Doctor/-Pope (7, resolved to `C4a`/`C5a`/`C4b` — each inheriting its base
category's Ant2/Versum2 with no override needed), several confessors + apostles-plural (2), and 5
fully-proper Lord's/BVM feasts given their own real `[Ant 2]`/`[Versum 2]` straight from their
Sancti files. **213 of ~218 real saints covered** at that point
(`tests/propers/commemoration-sweep.test.ts`, 213 cases); every entry not independently
live-verified marked `status: draft`.

The last 5 (the Vigils, `common-of-a-vigil`) looked unsourceable at first — no dedicated "Commune
Vigiliarum" antiphon set exists in the reference engine. The user pushed back explicitly ("are you
sure -- what happens if you pull the reference engine?") rather than accept that at face value.
Live-querying real commemorated instances (Vigil of St. Lawrence on a real Monday, Vigil of the
Assumption across 6 different real weekdays) found the actual mechanism behind each source file's
own `[Rule]: Versum Feria` line: a mere commemoration doesn't reuse the vigil's own antiphon at
all — it borrows the plain ferial day's antiphon/versicle instead. The antiphon quotes the
Benedictus canticle in weekday sequence (same shape `vespers-magnificat-antiphons.yml` already
uses for the Magnificat), the versicle is a single fixed text
(`lauds-versicle-monday.yml`'s own), and Sunday genuinely has no entry — a Vigil transfers off an
ordinary Sunday rather than ever reaching a commemoration there, confirmed live and already true in
this app's own precedence rules. Implemented as a small weekday-aware branch in
`sanctoralCommemorationPart` (now takes `weekday`) rather than a per-saint file.

**Full 218/218.** New/changed tests: `tests/propers/commemoration-sweep.test.ts` (213 static
bundles), `tests/hours/vigil-commemoration.test.ts` (3 cases, the dynamic weekday mechanism),
`tests/hours/lauds.test.ts`'s new Felix/Adauctus case. 176 new `*-commemoration.yml` files plus
one data file (`lauds-vigil-commemoration-antiphons.yml`). `npm test` (774 passed) and
`tsc --noEmit` pass.

### Vespers Magnificat antiphon — the last two cases closed as not gaps (2026-08-30)

Follow-up the same day: asked what the two remaining Magnificat gaps from the entry above actually
were (Advent 4's own Sunday landing on Dec 24; `marian-saturday`), then pushed back on "there's
nothing authored, but do you need to author content that never happens?" — a fair question the
TODO list hadn't actually answered. Checked both directly rather than leave them as
"unauthored, low priority": the Vigil of Christmas never has its own Vespers in the real rite at
all (Dec 23 evening is plain Advent-4 ferial Vespers with the O-Antiphon override, not First
Vespers of the Vigil — checked live; Dec 24 evening is always Christmas's own First Vespers —
"Vespera de sequenti; nihil de præcedenti," the reference engine's own words), and
`marian-saturday`'s own Commune file (`C10`) has no `[Ant Vespera]` tag at all, because a
Saturday's Vespers is always structurally ceded to the following Sunday under the real rubrics,
not just under this app's implementation of them.

Neither is "unauthored content pending" — there is no historical text for either case to author.
Dropped from the "Open work" list entirely rather than marked done-with-a-caveat, since authoring
something anyway would mean inventing text that never existed.

### Bilingual header `nameLa` mechanism (2026-08-31)

Picked up the standing "Latin saint/octave/named-feast names" backlog item. The `Bi`
(`{en, la}`) *display* wrapper already existed in `calendar/day-label.ts`, but the underlying
*data* had no bilingual name field at all — `SaintRecord.name`, `TemporalFeastRecord.name`, and
the derived `SanctoralIdentity`/`ActiveOctave`/`Commemoration` types were plain English strings,
so every proper name necessarily fell back through `biFallback`.

Added an optional `nameLa?: string` alongside `name` on `SanctoralIdentity` (`calendar/types.ts`),
`SaintRecord`/`SanctoralCandidate` (`calendar/feasts.ts`), `TemporalFeastRecord`
(`calendar/temporal-feasts.ts`), and `ActiveOctave` (`calendar/octaves.ts`), then threaded it
through every construction site that had been spreading `id`/`name`/`rank` (`collision.ts`,
`commemorations.ts`, `movable-feasts.ts`, `index.ts` x3, `vespers.ts` x2,
`octaves.ts`'s `considerCandidate`). `day-label.ts`'s old `biFallback(name)` calls became
`biName(name, nameLa)`, which still falls back to the English string when `nameLa` is absent —
the "surface real text, don't block on missing" convention preserved exactly, just now backed by
a real field instead of no field at all.

Authored `nameLa` for all 11 named temporal feasts as the proof set (each sourced from the
reference engine's own `[Officium]`/`[Rank]` title, adapted from the source's genitive/ablative
liturgical-title style to this app's nominative-noun-phrase naming convention — e.g. "In
Ascensione Domini" → `Ascensio Domini`, "Sacratissimi Cordis Domini Nostri Jesu Christi" →
`Sacratissimum Cor Domini Nostri Jesu Christi`) plus St. Lawrence as the one saint, chosen
specifically to exercise the octave-headline (`octaveCoreName`) and octave-commemoration code
paths, not just the plain-winner path the temporal feasts alone would have covered. Live-verified
end-to-end via a throwaway scratch test (removed after confirming) that Pentecost's/the
Ascension's/the Immaculate Heart's own winner label, St. Lawrence's mid-octave and closing-day
octave label, and his octave *commemoration* line (Aug 12) all render the authored Latin, not the
English fallback. Six pre-existing `toEqual` test fixtures across `august-sanctoral.test.ts`,
`octaves.test.ts`, `transfer.test.ts`, and `december-sanctoral.test.ts` needed their expected
objects updated to include the new `nameLa` key (Christmas's own octave-commemoration entries too,
since `christmas-day.yml` picked up a name in this same pass). `npm test` (807 passed) and
`tsc --noEmit` pass.

**Bulk pass, same day**: authored `nameLa` for the remaining 299 real saints (every
`src/data/calendar/saints/*.yml` except `example-confessor`, a deliberate non-real placeholder
never wired into `sanctoral-calendar.yml`). Sourcing standard for this bulk pass was lighter than
per-saint reference-engine verification (infeasible at this volume in one sitting): standard,
well-attested ecclesiastical Latin forms — the saint's own canonical Latin name plus a Latin
rendering of their English descriptor (`Bishop` → `Episcopus`, `Martyr` → `Martyr`, `Virgin` →
`Virgo`, `Confessor and Doctor of the Church` → `Confessor et Ecclesiæ Doctor`, etc.), nominative
noun phrase throughout, matching the temporal-feast proof set's own convention. Multi-saint feast
names (e.g. `ss-perpetua-and-felicity`) got the equivalent Latin plural construction
(`Sanctæ ... Martyres`). This is a different, lower sourcing bar than the rest of this app's
content (which is reference-engine-verified line by line) — flagged here explicitly since a
future pass may want to spot-check or upgrade individual entries against the reference engine's
own titles, the way the 11 temporal feasts and St. Lawrence were.

Applied via a scratch Python script (`nameLa` inserted right after each `name:` line) plus two
follow-up passes fixing ~180 pre-existing test fixtures across 18-20 test files that asserted
exact `toEqual` objects on `day.winner`/`day.commemorations` (now carrying the new `nameLa` key).
One real mechanism bug surfaced and fixed by this: `calendar/index.ts`'s sanctoral-vs-sanctoral
collision-resolution branch (`resolveDay`'s per-day candidate loop, ~line 40) was rebuilding the
winning candidate as `{ id, name, rank }` without `nameLa`, silently dropping it for any saint
that had *beaten another candidate on the same date* even when winning outright — caught by
`st-giles`, `st-justina`, `nativity-bvm`, and `vigil-of-st-lawrence` all resolving `nameLa:
undefined` despite having real authored data. Also found and fixed one stale test expectation
(`day-label.test.ts`'s Trinity Sunday/Queenship-of-the-BVM case) that had hardcoded "no Latin
name yet" as the expected behavior — now asserts the real Latin. `npm test` (807 passed) and
`tsc --noEmit` pass throughout. **Closes this backlog item** — moved out of the "Open work"
priority list.

### Vespers Magnificat antiphon — sanctoral Common-category fallback, 11/11 (2026-08-31)

User flagged today's (Aug 31, St. Raymond Nonnatus) Vespers was missing its Magnificat antiphon.
Root cause was systemic, not date-specific — the earlier 2026-08-30 pass (see above) had closed
the *temporal*-cycle gap in `getMagnificatAntiphon`, but the *sanctoral* Common-category
fallback (`magnificat-antiphon-${saint.benedictusCommon}`) had never been authored at all: only
11 saints ever need it (the ones with `propers: null` or no own antiphon, falling back to
`benedictusCommon`'s 11 distinct category values), and every single one was `missing`.

Sourcing method: each Commune source file (`Latin/Commune/C*.txt`) lays out
Vespers1(`[Ant 1]`) -> Matins -> Lauds(`[Ant 2]`) -> minor hours -> Vespers2(`[Ant 3]`), confirmed
directly by live query (Ss. Placid and Companions, 10-05-2026: the evening before renders
`[Ant 1]` under "Vespera de sequenti", the feast's own evening renders `[Ant 3]`) and
cross-checked against 3 more saints (St. Blaise, St. John Gualbert). Authored 10 of 11 categories'
`magnificat-antiphon-common-of-*.yml` from each Common's own `[Ant 3]`
(`common-of-a-confessor`/`common-of-an-abbot` share one file, C5, so their Magnificat text is
identical by design, same as their existing Benedictus text).

Found and fixed two pre-existing bugs in sibling `benedictus-antiphon-common-of-*.yml` files
while cross-checking `[Ant 2]` against them: `common-of-a-confessor` and `common-of-the-bvm` had
both been sourced from `[Ant 1]` (First Vespers) instead of `[Ant 2]` (Lauds) — live-verified
wrong via St. Alexius's real Lauds query (07-17-2026, renders "Euge, serve bone", not the
file's old "Similabo eum"). Corrected both.

**The 11th category, `common-of-a-vigil`, needed a code fix instead of a content file.**
Initially assumed unreachable (a Vigil's own *Second* Vespers always loses to the following
feast's First Vespers — true, and confirmed via `compareFeastClass('vigil', ...)` always losing
in `resolveEveningDay`). But a Vigil's own *First* Vespers (anticipated the evening before, since
`hasFirstVespers` has no rank floor) absolutely does reach this code path — caught by re-checking
after the user asked "what about 11 of 11?". Live-verified across 4 different years
(2025-2028) that the real office never gives a Vigil a fixed proper Magnificat text at all — its
`[Rule]` block ("ex C1v; ... Laudes 2") always borrows whatever the current ferial weekday's own
antiphon is (`{Antiphona ex Proprio de Tempore}` every single year tried). So `common-of-a-vigil`
correctly has no content file; instead `getMagnificatAntiphon` (`hours/resolve-common.ts`) now
special-cases `saint.rank === 'vigil'` to fall through to the same `vespersMagnificatAntiphons`
weekday-default table the plain temporal branch already used, before ever trying a Common
lookup. `npm test` (810 passed) and `tsc --noEmit` pass.

### Ferial Matins nocturn: one antiphon per psalm (2026-09-01)

User flagged that the ferial (1-nocturn) weekday antiphons looked inconsistent — some psalms
paired under one shared antiphon, others solitary, within the same nocturn. Root cause: the
source's own per-weekday tables (`Daym1_`-`Daym6_`) do frame several *different-psalm* pairs
under one shared antiphon (normal Monastic practice), and vu's `matins-ferial-antiphons.yml`
had carried that pairing straight through — reading fine in the source's own untouched weekly
layout, but looking arbitrary once `psalter-distribution.yml`'s reslice left some of those pairs
intact and split others across weekday boundaries. Fixed by splitting every cross-psalm pair
into two single-psalm groups (a same-psalm split across a chapter boundary, e.g. Ps 36:1-26/
36:27-40, is unaffected — that's still one antiphon per psalm).

For each split pair, whichever psalm the antiphon's own words actually quote (checked against
`src/data/psalms/*.yml`, not assumed from position — several pairs listed the *second* psalm's
own words under the first psalm's slot, e.g. Wednesday's "Quam bonus Israël Deus" pair, [71,72],
is actually Ps 72's own opening, not 71's) keeps that antiphon; its former partner got a new
one, per direct instruction: search the reference corpus for a real one before inventing.
Sourcing that produced (6 of 10 real, 4 composed fresh — all directly quoting that psalm's own
Vulgate text either way, the same convention every other antiphon in this corpus follows):
- **Real, borrowed**: Ps 39 and 71 from `Tempora/Quad6-5` and `Tempora/Quad6-4` (Holy Week
  ferial propers); Ps 55 from `Tempora/Quad5-5` (Lenten Friday ferial); Ps 102 from
  `Tempora/Pent01-4`; Ps 107 from `Tempora/Pent02-5` (post-Pentecost ferial propers); Ps 48 from
  the Roman *secular* ferial Matins table (`[Day0]`-`[Day6]`, same file as the Monastic table,
  individually-antiphoned rather than paired — see [[vu_secular_source_ok]]).
- **Composed fresh** (no match found anywhere in the corpus): Ps 33, 43, 79, 83 — each from its
  own opening verse, in the same incipit style the rest of the corpus already uses.

`tests/hours/matins.test.ts`'s ferial test updated (was asserting "antiphon only on each group's
first psalm" — now every psalm in the nocturn has one). `npm test` (814 passed) and
`tsc --noEmit` pass.

**Not done in the same pass, confirmed as a separate real gap**: Sunday Matins' Nocturn 3
"Allelúja, allelúja, allelúja" antiphon is wrong for Septuagesima-Passiontide — see this file's
"Mechanism gaps" section above for the live-verified detail and the 9 antiphons already
collected.

### Sunday Matins Nocturn III seasonal override, Septuagesima-Palm Sunday (2026-09-01)

Fixed the gap flagged above. New `data/hours/matins-sunday-nocturn3-overrides.yml`, keyed by
`temporalId`, plus `hours/matins.ts`'s `sundayNocturn3For` (looked up with the `temporalId`
`resolveOrdo` already computes, merged field-by-field over `matins-sunday-antiphons.yml`'s base
`nocturn3` group — only the real Sunday branch uses it, not the Duplex+-weekday-feast branches,
which stay saint/Common-specific as before).

Re-verifying live (one `prayMatutinum` query per Sunday of the 2026 cycle, Monastic Tridentinum
1617) turned up more than the antiphon-only change the earlier note assumed:
- **Septuagesima, Sexagesima, Quinquagesima**: antiphon only — canticles (Isa 33:2-10/33:13-18,
  Sir 36:14-19) and versicle ("Exaltáre Dómine in virtúte tua. / Cantábimus et psallémus
  virtútes tuas.") are the same as the plain-season default.
- **Lent 1-4**: a genuinely different set of 3 canticles (Jer 14:17-21, Lam 5:1-7,15-17,19-21,
  Ezek 36:24-28 — new `data/scripture/jer-14.yml`, `lam-5.yml`, `ezek-36.yml`, the first content
  under the `lam` book code) *and* a different versicle ("Scuto circúmdabit te véritas ejus. /
  Non timébis a timóre noctúrno."), on top of the antiphon.
- **Passion Sunday, Palm Sunday**: base canticles, but the versicle is "Ne perdas cum ímpiis,
  Deus, ánimam meam. / Et cum viris sánguinum vitam meam." — not "Exaltáre Dómine." The earlier
  note's placeholder Passion Sunday antiphon text ("Deus ádjuvat me...") was actually this
  versicle, not the antiphon; the real antiphon is "Ego glóriam meam non quæro: † est qui
  quærat, et júdicet." (John 8:50, that Sunday's own Gospel). Palm Sunday's full antiphon:
  "Hosánna fílio David: † benedíctus, qui venit in nómine Dómini, Rex Israël: † Hosánna in
  excélsis."

New regression tests in `tests/hours/matins.test.ts` cover all three shapes (antiphon-only,
antiphon+canticles+versicle, antiphon+versicle-only). `npm test` (817 passed), `tsc --noEmit`,
and `npm run build` all pass.

### Advent Sunday Matins Nocturns I-III seasonal override (2026-09-01)

Follow-up, same day: the user asked whether Advent Sundays' own Nocturn III needed the same kind
of fix. Live re-check (`prayMatutinum`, Monastic Tridentinum 1617, all 4 Advent Sundays of the
2026 cycle) found the answer is yes, and bigger than just Nocturn III — Advent's own seasonal
antiphons apply to **all three nocturns**, not just Nocturn III, and identically across all 4
Sundays (no per-Sunday variation the way Septuagesima-Palm has).

- **Nocturns I-II**: psalm numbers unchanged (Ps 20-31), but every one of the 6 antiphons and
  both versicles are seasonal — none of which existed anywhere in the codebase before this fix
  (unlike Nocturn III, Nocturns I/II had *no* override mechanism at all; `matins.ts` used the
  plain-season `matins-sunday-antiphons.yml` base unconditionally).
- **Nocturn III**: swaps the plain-season 3-canticle set (Isa 33:2-10/33:13-18, Sir 36:14-19) for
  a wholly different one (Isa 40:10-17, 42:10-16, 49:7-13 — 3 new `data/scripture/isa-{40,42,49}.yml`
  files), under its own Marian antiphon ("Gábriel Ángelus locútus est Maríæ...") and versicle.

New `data/hours/matins-sunday-advent-nocturn-overrides.yml` — a single season-wide block (keyed
by season, not per-Sunday `temporalId`, since all 4 Advent Sundays are identical) — plus
`hours/matins.ts`'s generalized `sundayNocturnFor(nocturn, day, temporalId)`, which now checks
Advent's season-wide override ahead of (Nocturn III only) the existing per-Sunday Septuagesima-
Palm `sundayNocturn3For` lookup; Advent and Septuagesima-Palm never overlap so there's no
precedence conflict between the two override tiers.

New regression test in `tests/hours/matins.test.ts` (`Advent Sunday Nocturns I-III seasonal
override`) asserts all 4 Advent Sundays render identically. `npm test` (818 passed),
`tsc --noEmit`, and `npm run build` all pass.

### Paschaltide Sunday Matins Nocturns I-III seasonal override (2026-09-01)

Second follow-up, same day: does any other season need the same treatment? Live sweep of the
reference source (`Psalterium/Psalmi/Psalmi matutinum.txt`) confirmed only 3 season-tagged
blocks exist at all — `[Daym0]` (plain default), `[Advm0]` (Advent, done above), `[Paschm0]`
(Paschaltide) — so Christmastide and post-Epiphany Sundays have no gap; Paschaltide does.

Ordinary Paschaltide Sundays (Low Sunday through the Sunday before Ascension — `day.season ===
'eastertide'`) get the same seasonal antiphon/versicle/canticle treatment as Advent, live-
verified identical across 5 Sundays of the 2026 cycle (04-12 through 05-10). New Nocturn III
canticles: Isa 63:1-5, Osee 6:1-6, Sophonias 3:8-13 (3 new scripture files — first content
under the `osee`/`soph` book codes). Structurally different from Advent's own shape, not just
different text: Nocturns I and II each take a **single** antiphon covering all 6 psalms of the
nocturn, not 3 separate antiphons (one per psalm-pair) — confirmed by the live query showing
only one antiphon before/after each whole nocturn's psalmody. Nocturn III's own antiphon is
unchanged from the plain-season "Allelúja, allelúja, allelúja" default; only its canticles and
versicle differ.

New `data/hours/matins-sunday-paschaltide-nocturn-overrides.yml`; `hours/matins.ts`'s
`sundayNocturnFor` generalized to a season-keyed lookup (renamed `MatinsSundayNocturnOverrides`)
covering both Advent and Paschaltide.

**Explicitly NOT covered by this fix, each its own separate thing**:
- **Easter Sunday itself** — has a wholly proper Matins (different psalms too, not just
  antiphons); this is the already-tabled "Easter's own octave" mechanism gap above, not folded
  into this seasonal-overlay shape.
- **The Sunday within the Octave of Ascension** (`ascensiontide`'s one Sunday) — live-checked
  (2026 proof date 2026-05-17) and found to render its own genuinely proper content ("Eleváta
  est...", Ps 8 instead of the fixed Ps 20-31 scheme) in the reference engine, because that
  date's real office winner is a Duplex-rank saint (St. Paschal Baylon) whose own Common/proper
  Matins wins outright. **Investigated, not a bug**: this app's Sunday psalmody is documented
  above (Priority section, "Sunday's fixed 12-psalm/3-canticle psalmody") as deliberately fixed
  and unconditional on every Sunday regardless of the office winner — a direct-instruction
  non-reconstruction choice, not an oversight. The winning saint's own content still surfaces
  generously through the separate Nocturn 2/3 *readings* pool (`buildReadingPool`, already
  saint-aware); only the antiphon/psalm layer stays fixed. Changing the branch dispatch to let
  a Sunday's winner override the fixed psalmody would reverse a settled design decision, not
  close a gap — left alone.

New regression test in `tests/hours/matins.test.ts` (`Paschaltide Sunday Nocturns I-III
seasonal override`) asserts all 5 ordinary Paschaltide Sundays render identically. `npm test`
(819 passed), `tsc --noEmit`, and `npm run build` all pass.

### Sunday within the Octave of Ascension: real precedence bug, fixed (2026-09-01)

User pushed back on the "not a bug" conclusion above: correctly — the "Duplex+ saint wins
outright over an Ascensiontide Sunday" case was a real precedence bug, not settled design.
`data/calendar/temporal-categories.yml` had `ascensiontide`'s Sunday coded `ordinary-sunday`
(Duplex+ wins outright), the only "great octave" season not already `privileged-sunday`
(eastertide/pentecost both already were) — no verification comment on that line, unlike every
other entry in the file. Live-verified (proof date 2026-05-17, St. Paschal Baylon, real Duplex):
the reference engine titles the day "Dominica infra Octavam Ascensionis ~ Semiduplex Dominica
minor" and gives it its own wholly proper Matins, no trace of the saint. Fixed: `ascensiontide`
is now `privileged-sunday`, matching eastertide/pentecost — St. Paschal Baylon now transfers
(`decideOccurrence`'s existing generic privileged-sunday branch handled this correctly with no
new code, just the corrected category).

With the calendar now resolving the *right* winner (the Sunday itself, `sunday-after-ascension`),
Matins still needed real content: this Sunday's own proper is a wholly different psalm scheme
from the fixed 12-psalm base — Ps 8/10/18/20/23/29 for Nocturn I, Ps 45/46/95/96/98/102 for
Nocturn II, each of the 12 psalms individually antiphoned (not paired) — plus its own versicles.
Nocturn III reuses the same 3 canticles as the ordinary-Paschaltide override (Isa 63:1-5, Osee
6:1-6, Sophonias 3:8-13, already authored) under its own antiphon ("Illi autem profécti
prædicavérunt ubíque...") — English for that one antiphon is this app's own translation (the
reference engine's own English tree is genuinely blank there, confirmed not a fetch gap), since
it's an unambiguous, well-known formula (Mark 16:20).

New `data/hours/matins-sunday-named-nocturn-overrides.yml` — a per-`temporalId` override layer
(currently just `sunday-after-ascension`) for a specific privileged Sunday whose own proper is
wholesale different, distinct from the season-wide Advent/Paschaltide overrides above.
`hours/matins.ts`'s `sundayNocturnFor` now checks this map first (most specific), ahead of the
season-wide overrides, ahead of the plain default.

New regression tests in `tests/hours/matins.test.ts` (`Sunday within the Octave of Ascension`)
assert both the calendar winner (Sunday over the saint) and the full Matins content. `npm test`
(821 passed), `tsc --noEmit`, and `npm run build` all pass.

### Full sweep of `temporal-categories.yml` — two more real bugs found and fixed (2026-09-01)

Follow-up, same day: user pushed back hard on the "not a bug" framing above ("the immaculate
conception transfer is correct... annuncition should transfer") and asked to sweep every
remaining unverified entry in `data/calendar/temporal-categories.yml`, not just the one already
fixed. Found two more real, calendar-wide bugs:

1. **`commemorations.ts`'s `privileged-sunday` case itself was wrong**, not just one season's
   tag. It gave any duplex-majus+ candidate a bare commemoration instead of transferring —
   never actually verified against a primary source. Live-checked 3 independent cases: the
   Immaculate Conception (Duplex I. classis) transfers off Advent II with **no** commemoration
   (2024/2030/2041); the Annunciation (also Duplex I. classis) transfers off both Palm Sunday
   and Easter Sunday itself the same way (2029/2035/2040/2046). Confirmed by direct instruction
   as correct (with a known follow-on: the Annunciation's *real* rule jumps it past the whole
   Easter Octave rather than the next open day, which lands inside it — blocked on the
   not-yet-built Easter-octave mechanism, noted but not built here). Fixed: `privileged-sunday`
   now transfers every rank unconditionally, no exception.

2. **`eastertide`'s own `sunday` tier was backwards** — coded `privileged-sunday`, should be
   `ordinary-sunday` for the *ordinary* Sundays after Easter (not Easter Day itself, not within
   its Octave, both already separately handled). Live-verified twice: the Finding of the Holy
   Cross (Duplex-II-classis, 2026-05-03) and St. Mark (Duplex-II-classis, 2027-04-25) both win
   outright over "Dominica IV Post Pascha," with the Sunday merely noted as a `Tempora` line
   underneath — exactly `ordinary-sunday`'s rule. Easter Day itself still needs the stronger
   tier (confirmed by the Annunciation evidence above), so it's now special-cased directly in
   `calendar/temporal.ts`'s `resolveTemporalCategory` (Sundays never consult `offsets`, so this
   couldn't be expressed as a plain offset entry the way every other exception in that file is).

**Six test failures from the `privileged-sunday` fix, all genuine fallout, not regressions** —
each test had encoded the old, wrong "commemorated in place" behavior as its expected value:
St. Andrew (Duplex II. classis) no longer commemorated alongside Advent I, but shown
transferred away (landing on Dec 1, where he now wins outright — `privileged-feria-minor`'s
Semiduplex+ threshold easily clears); St. Antoninus similarly now wins outright on his own day
(the real Easter-6 Sunday) instead of being pushed to the next day by the old wrong rule. All
six tests updated to real, live-verified dates/values, not just loosened to pass.

**Remaining entries in the file, not independently live-collision-tested but documented as
such** (each has its own comment now): the Vigil of Pentecost and Vigil of Christmas
(`privileged-feria-major`) are corroborated by their own formal rank in the reference source
("Semiduplex I classis" / "Duplex I classis" respectively) rather than a live saint collision,
which is rare by construction; Pentecost Sunday itself (`privileged-sunday`) is consistent with
the confirmed "great octave Sunday" pattern but has no fixed-date-saint collision to test
against; Corpus Christi/Sacred Heart's `feria` (`privileged-feria-major`) shares the same
"feast of the Lord" reasoning already confirmed elsewhere, and their `sunday` key is flagged as
genuinely unreachable (both are always fixed weekdays, never a Sunday, per
`easter-offsets.yml`'s own `days:` table). Every other `feria` entry is the trivial
`ordinary-feria` default (no rank threshold — any real content wins), already implicitly
exercised correctly across hundreds of ordinary weekdays in the existing test suite.

`npm test` (820 passed — one net fewer than before since two tests merged into one clearer
case), `tsc --noEmit`, and `npm run build` all pass.

### Easter octave transfer cascade — mechanism built (2026-09-01)

Closes the "Easter's own octave" item from this file's own Priority list. User asked for it
directly, then answered two design questions with real evidence rather than guesswork:
"what happens to the saint's own day" and "what are the gating rules" for something impeded
within Holy Week/the Easter Octave.

**Root cause, found by reading the existing code rather than assuming**: `resolveDay` only ever
checked yesterday/tomorrow for an inbound transfer — a single hop. There was already an honest,
existing comment marking this exact gap (`applyIncomingTransfer`'s `privileged-feria-major`
branch: "a transfer chain — not modeled... simply not delivered here rather than guessed at").
Anything impeded in Holy Week just vanished rather than reappearing.

**Live-verified real behavior (2033: Palm Sunday Apr 10 → Low Sunday Apr 24, five real saints
falling inside that window) before building anything**:
- A candidate impeded anywhere in the whole span (Holy Week, Easter Day, the Octave, Low
  Sunday) skips the *entire* span — not just the day it was impeded on — and keeps walking
  forward day by day past it, same as any ordinary transfer, until it finds a day that isn't
  already spoken for by a real native occurrence or an earlier-queued candidate ahead of it in
  line. St. Leo I (native Apr 11, Duplex) landed Apr 27; Ss. Soter & Caius (native Apr 22)
  landed Apr 28 (right after Leo); St. George (native Apr 23) had to wait until May 4, well
  past several already-occupied days.
- **Multiple candidates can be pending at once** — a real FIFO queue ordered by native date,
  not just "the one most recent candidate." All three above overlapped in their pending window.
- **A transferred-in candidate never gets the "duplex-majus+ still gets a nod" commemoration**
  a *native* occurrence of that rank would on one of these days — confirmed with the
  Annunciation (Duplex I. classis, native Easter Sunday itself in 2035): zero commemoration
  anywhere across the whole span, landing outright on Apr 4 (past Low Sunday, Apr 1 that year).
  This is genuinely different from a native duplex-majus+ candidate (e.g. St. Mark, native
  within some other year's Octave, which *does* get commemorated in place) — the "never
  displaced, but gets a nod" treatment is reserved for a day's own native candidate, not
  anything merely passing through.
- **Low Sunday itself needed the same fix as Easter Day** — found while building this: it
  carries the same formal "Duplex I. classis" rank as Easter Day in the source, but was
  falling through to eastertide's plain `ordinary-sunday` default (today's earlier fix only
  covered offset 0). Not independently collision-verified (no Monastic-track saint — a real
  `SanctiM/MM-DD.txt` file — ever lands there across the real range of Easter dates), but
  corroborated by its own rank label and by the complete absence of even a `Transfer:` note for
  a same-dated non-Monastic-track saint (St. Isidore of Seville 2027, St. Vincent Ferrer 2043)
  that a real Monastic day would show. Fixed in `calendar/temporal.ts`'s `resolveTemporalCategory`
  alongside offset 0.

**What got built**: a real `MAX_FORWARD_TRANSFER_LOOKBACK_DAYS`-bounded (60 days, generous
headroom over the worst observed real case) queue simulation, `resolveForwardTransferLanding`
in `calendar/index.ts` — walks chronologically from the lookback window up through the target
date, replaying each day's own native occurrence and, if a candidate is already queued, one
landing attempt per day (reusing `applyIncomingTransfer` completely unchanged — its existing
`privileged-feria-major` no-op turned out to already be exactly the right single-hop refusal;
the fix was purely wrapping it in an actual cascade instead of a dead end). Only forward-
direction transfers chain this way; a Vigil's backward transfer stays single-hop as before —
structurally different (displaces backward in time) and not observed to need it.

New tests in `tests/calendar/transfer.test.ts` cover the full chain (Leo/Hermenegild/Tiburtius
landing in strict native-date order past the whole 2033 span) and the Annunciation's own
zero-commemoration case (2035). `npm test` (823 passed), `tsc --noEmit`, and `npm run build`
all pass.

### Correction, same day: `privileged-feria-major`'s own "duplex-majus+ still gets a nod" rule was wrong

User pushed back immediately: "you dont commemorate -anything- from palm sunday through low
sunday" — and was right. The cascade mechanism above was built reusing `applyIncomingTransfer`'s
existing `privileged-feria-major` no-op unchanged, on the assumption that its sibling behavior
in `decideOccurrence` (a *native* duplex-majus+ candidate there gets a bare commemoration, only
below-majus ranks transfer) was itself correct — it had a citation (St. Mark, Duplex II. classis,
"merely commemorated" within the Easter Octave). That citation turned out to be a misreading,
not just unverified.

Live-checked two more real dates before touching code: the Annunciation (Duplex I. classis, the
highest rank that exists) transfers with **zero** commemoration on Holy Thursday itself in years
it lands there (2027/2032/2043) — not just off a Sunday, off an ordinary Holy Week feria too.
Then checked whether this was Holy-Week-specific or true of every `privileged-feria-major` use:
Chair of St. Peter at Antioch (Duplex majus) transfers the same way off Ash Wednesday (2034).
So the exception was wrong everywhere this category is used, not just the Easter span.

Fixed: `commemorations.ts`'s `privileged-feria-major` case now matches `privileged-sunday`
exactly — transfers every rank unconditionally, no commemoration ever, for anyone. Three
existing tests had the old (wrong) assumption baked into their expected values and needed
updating to real, live-verified outcomes, not just loosened: St. Mark (2030) now transfers
instead of being commemorated, landing May 5 that year (vu's own dense April/May calendar
pushes him well past his native date); the Annunciation's own 2035 landing moved from Apr 4 to
Apr 6, since two *earlier*-native-dated candidates (St. Benedict, also Duplex I. classis, native
Mar 21; St. Gabriel the Archangel, Duplex majus, native Mar 24) are now also correctly caught by
the same span that year and queue ahead of her.

`npm test` (822 passed), `tsc --noEmit`, and `npm run build` all pass.

### Easter-octave backlog redesigned — one-per-day, floor-gated, no rank fight (2026-09-01)

Same day, a real design conversation rather than another live-verification pass: user rejected
the multi-day cascading queue built above ("that's not a plan... we cant do a batch
commemorations. we can do one a day."). New shape, per direct instruction:

- **Gate**: only candidates at or above `duplex-majus` are worth carrying out of the span at
  all — below that, a candidate simply lapses for the year, no trace, same as it already gets
  none *within* the span itself. (Confirmed as the right floor by inventorying every fixed-date
  saint that could ever fall in the whole reachable window, March 15 - May 2: only 5 ever clear
  it — St. Benedict, St. Gabriel the Archangel, the Annunciation, St. Mark, Ss. Philip & James
  — everything else in that range is plain Duplex or below.)
- **Landing**: the Nth day after Low Sunday celebrates the Nth candidate that cleared the floor
  that year, in native-date order — winning outright, no rank comparison against whatever would
  natively be there, which is simply demoted to a commemoration. One candidate per day,
  positional, not a search for the next free day.

Replaced `resolveForwardTransferLanding`'s generic multi-day queue with two pieces:
`easterOctaveBacklogFor(year)` (computes that year's whole backlog once, filtered to the floor)
and `applyEasterOctaveBacklog` (the Nth-day lookup), both in `calendar/index.ts`. The ordinary
single-hop yesterday/tomorrow transfer check (used everywhere else) is back to its original
form — the whole redesign is now a dedicated, narrowly-scoped mechanism (same architectural
pattern as `applyMarianSaturday`/`applyAdventFourVigilCommemoration`), not a generalized queue.

**Real bug caught while re-verifying**: the ordinary single-hop check was independently picking
up Low Sunday's own transfer signal and landing/commemorating a below-floor candidate (St.
Fidelis, plain Duplex) the very next day — bypassing the new floor entirely. Fixed by excluding
the whole Palm-Sunday-through-Low-Sunday span from that ordinary path (scoped by `easterOffsetOf`,
not by category, so it doesn't also swallow an unrelated privileged Sunday elsewhere in the year
like Advent II).

Rewrote the tests this broke to real, live-verified outcomes for the new shape (2033: nothing
that year clears the floor, so the whole backlog is empty and every impeded candidate lapses;
2035: three candidates clear it and land the 1st/2nd/3rd day after Low Sunday in order). Then,
having inventoried the 5 always-reachable duplex-majus+ candidates together, one more direct
instruction: St. Patrick's own rank is bumped from the source's plain Duplex to duplex-majus
(`data/calendar/saints/st-patrick.yml`) — a deliberate departure, not a transcription, same
category of move as this app's existing late-canonization rank-deflation convention, just in
the other direction — bringing his own count to 6.

`npm test` (821 passed), `tsc --noEmit`, and `npm run build` all pass.

### Matins ferial nocturn: mid-nocturn versicle removed, a real live-verified bug (2026-09-02)

User noticed the 1-nocturn ferial branch had a V./R. versicle stuck mid-list as if marking a
nocturn boundary — structurally incoherent for a single continuous nocturn.

`data/hours/matins-ferial-antiphons.yml`'s per-weekday `groups` carried a `versicle` field
ported from a mid-list V./R. line embedded in the source `Psalmi matutinum.txt`
(`[Daym1_]`-`[Daym6_]`), rendered partway through the single continuous 1-nocturn ferial day.
Live-verified against the reference engine (`Pofficium.pl`, `command=prayMatutinum`,
`version=Monastic Tridentinum 1617`) for Wednesday and Thursday: the real office runs straight
from the last psalm to the Capitulum, no versicle anywhere in the psalmody — that source line
is inert there, never actually emitted. Removed the field from all 6 weekdays and the
rendering code (`ferialAntiphonedNocturn`); the earlier "versicle placement" reasoning in that
file's header was a documented-but-unverified assumption, not real content.

See memory `vu_ferial_matins_antiphons` for full detail.

`npm test`, `tsc --noEmit`, and `npm run build` all pass.

### Matins: 3-nocturn threshold lowered from Duplex to Semiduplex (2026-09-02)

Same session, a separate fix: the 3-nocturn gate (`hours/matins.ts`'s `threeNocturns`) was
floored at Duplex (`isDoubleOrHigher`). User's own design call, prompted by noticing today's
Matins (St. Stephen of Hungary, Semiduplex) wrongly rendering as 1-nocturn: Semiduplex and
Duplex should share the same 3-nocturn/3-psalms-each shape (distinct from Sunday/Duplex-majus
+'s 6/6/3-plus-canticles scheme). Lowered the floor with a new, separate check
(`isAtLeast(winner.rank, 'semiduplex')`) rather than touching `isDoubleOrHigher` itself, which
drives an unrelated antiphon-doubling rule elsewhere and must stay floored at Duplex. No new
content authoring was needed — `ferialPsalmodyThreeNocturns` already existed and already
chunks the ferial 9-psalm pool into 3/3/3 as the fallback for a plain-Duplex winner with no
override authored; Semiduplex winners now fall into that same already-working branch. Proved
via St. Stephen of Hungary (2026-09-02, Common of a Confessor): 3 real nocturns, Common's own
Terce/Sext/None antiphons+versicles, Te Deum present.

See memory `vu_matins_semiduplex_threshold` for full detail.

`npm test` (823 passed), `tsc --noEmit`, and `npm run build` all pass.

### Easter's own octave — Sunday/Monday/Tuesday proper Matins authored, general temporal-override mechanism added (2026-09-02)

Closes most of the long-tabled "Easter's own octave" mechanism gap (Priority section, item 1).
User pushed back on an initial overcomplicated read ("psalms with lections" looked like a new
rendering shape at first glance) — turned out, after live-verifying the actual structure, to
be exactly what the user guessed: a temporal override file that "looks weird" only because of
its content (13 psalms under one antiphon per nocturn, not the usual per-pair grouping), not
because it needs new mechanism. Every psalm already gets its own Gloria Patri centrally
(`hours/index.ts`'s `withGloriaPatri`) regardless of hour, and the 12 patristic lessons were
*already* authored (`data/propers/nocturn-readings/easter-sunday.yml`, from the earlier
temporal nocturn-readings sweep) — the only real gap was the psalmody/antiphon/versicle/
canticle layer.

**Content**: new `easter-sunday` entry in `matins-sunday-named-nocturn-overrides.yml` —
live-verified against the reference engine (`Pofficium.pl`, `prayMatutinum`, Monastic
Tridentinum 1617, 2026-04-05/06/07) — Nocturn 1: psalms 1,2,8,15,23,27 under one shared
antiphon ("Ego sum qui sum"); Nocturn 2: psalms 29,63,65,75,87,107 under one shared antiphon
("Terra trémuit"); Nocturn 3: the same 3 canticles already authored for the ordinary-
Paschaltide seasonal override (Isa 63:1-5, Osee 6:1-6, Soph 3:8-13), under its own antiphon
("Nolíte expavéscere"). All 3 nocturn-closing versicles are byte-identical to the ones
already authored in `matins-sunday-paschaltide-nocturn-overrides.yml`, confirming those are
shared season-wide rather than coincidentally reused.

**Mechanism (small, general-purpose, not Easter-specific)**: `hours/matins.ts`'s
`threeNocturns` gate and the Sunday-branch dispatch previously only ever fired for
`day.weekday === 'sunday'` or a sanctoral winner — a *temporal* winner had no path into 3-
nocturn proper content at all, even with a matching `sundayNamedOverrides` entry. Generalized:
a temporal winner with an authored named override is now eligible too (`hasTemporalNamedOverride`),
reusing the exact same `sundayPsalmNocturn`/`sundayCanticleNocturn` render functions the Sunday
and saint-proper branches already share — deliberately built as a reusable feature (user,
2026-09-02: "we want temporal override anyway... there are temporal feasts that could have
their own psalms/lessons"), not a one-off hack.

**The scoping wrinkle that made this non-trivial**: `winner.id` is `'easter-sunday'` for the
*entire* Octave (all 6 weekdays), but only Sunday/Monday/Tuesday are live-verified to share
this content — Wednesday-Saturday have their own different (Paschal-alleluia-flavored, not yet
authored) antiphons over a differently-shaped psalm set. A naive "any day with this winner.id"
check would have wrongly applied Sunday's content through Saturday. Fixed with a new optional
`weekdays` field on a named-override entry (`easter-sunday: weekdays: [sunday, monday,
tuesday]`) — omitted (as `sunday-after-ascension` still does) means "every day carrying this
temporalId," the common case for a temporalId that already maps to exactly one calendar day.

**Still open, a separate smaller gap**: Wednesday-Saturday of the Easter Octave — live-verified
(2026-04-08) to carry real Paschal-alleluia antiphons ("Jesum qui crucifíxus est quǽritis,
allelúja...") over a psalm set the reference engine sizes differently from vu's own
already-fixed 9-psalm ferial table (itself deliberate — see "No RB summer contraction" in
CLAUDE.md). 4 more days of antiphon-only content, same shape as the existing Advent/Paschaltide
seasonal-overlay pattern, layered onto vu's own ferial psalm numbers rather than trying to
reproduce the reference engine's differently-sized set — not started.

`npm test` (828 passed), `tsc --noEmit`, and `npm run build` all pass.

### Easter's own octave — Wednesday-Saturday's antiphons done too, fully closing the item (2026-09-02, same day)

Closes the remaining open piece from the entry above. Live-verified: all 4 weekdays
(Wednesday-Saturday) share the exact same 2 antiphons verbatim — not per-day content, despite
each drawing on a different Gospel reading in the nocturn-readings pool ("Jesum qui crucifíxus
est quǽritis, allelúja..." / "Veníte et vidéte locum ubi pósitus erat Dóminus, allelúja,
allelúja.", Pofficium.pl, prayMatutinum, Monastic Tridentinum 1617, 2026-04-08/09/10/11).

New general-purpose mechanism, the ferial-side counterpart to the Sunday-named-override built
earlier the same day: `data/hours/matins-ferial-named-antiphon-overrides.yml` (a temporal id's
own antiphon list, optionally restricted to specific weekdays — same `weekdays` idiom as the
Sunday-side override) + `hours/matins.ts`'s `ferialNamedAntiphonedNocturn`, checked first in
`ferialAntiphonedNocturn` ahead of the plain per-weekday default. The antiphon list is
distributed evenly across the weekday's own already-fixed psalm-group count (not tied to
specific psalm numbers), since one entry is meant to be reused across several different
weekdays rather than authored per-day — psalm numbers themselves are unchanged from the plain
ferial default (vu's own fixed table, not the reference engine's differently-sized Paschaltide-
contracted one).

With this, "Easter's own octave" is fully closed: calendar precedence (2026-09-01), Sunday/
Monday/Tuesday's shared proper psalmody (2026-09-02), and now Wednesday-Saturday's antiphons
(2026-09-02, same day). See memory `vu_easter_octave_proper_matins` for the full writeup.

`npm test` (832 passed), `tsc --noEmit`, and `npm run build` all pass.

### Bulk Vulgate/Douay-Rheims scripture import — completed from the real source, plus fixes upstream of it (2026-09-03)

Closes the long-deferred "bulk Vulgate + Douay-Rheims scripture import" item. What actually
happened, in order:

1. **Stale-status bug**: `src/propers/bible-plan.ts`'s `resolveReading` preferred each
   bible-plan YAML row's own hardcoded `status: { la: missing, en: missing }` field over
   computing status from the actually-resolved scripture text. That field was accurate before
   a same-day bulk import landed real text into `src/data/scripture/`, but was never cleared —
   so every Matins bible-plan reading still showed "translation pending" regardless. Fixed by
   stripping the stale field from all 368 `data/hours/bible-plan/*.yml` files.
2. **Book-abbreviation mismatch**: the bulk-imported store keys books by Vulgate-numbering
   abbreviation (`sap`, `jac`, `joan`, `1reg`..`4reg`, `1par`/`2par`, `1esdr`/`2esdr`, ...),
   matching `octave-readings.ts`'s pre-existing convention, while `bible-plan.ts`'s own TSV
   import used common English abbreviations (`kgs`, `sam`, `chr`, `john`, `jas`, `luke`,
   `mark`, ...) for the ~30 books where the two conventions diverge. Fixed with a
   `BOOK_ALIASES` map in `src/scripture/index.ts`, at the lookup layer, so bible-plan's
   citations keep their familiar English display form.
3. **Verse-number prefixes**: `resolvePassages` (shared by octave-readings/bible-plan/nocturn-
   readings) was prepending each verse's number to the joined reading text ("1 Et factum
   est..."), visible in rendered readings. Dropped — join verse text alone.
4. **3 books the bulk import missed entirely**: Jude, Obadiah, and Philemon (each a single
   short chapter) had no file at all in `src/data/scripture/`, unlike everything else the bulk
   import covered — hand-authored first, then replaced with source-derived text once the real
   source turned up (see next point) — small inaccuracies in the hand-typed version confirmed
   the value of going back to source.
5. **The rest of the gap, and the real source**: comparing every book/chapter citation used
   anywhere in the app against the store turned up 137 missing chapters, scattered oddly across
   ~30 otherwise-covered books (10 of Genesis's 50 chapters, 9 of Daniel's 14, etc.) — too much
   to hand-author reliably. The user located the original "Unbound Bible" source archives
   (`~/opt/inactive/bible/versions/douay_rheims.zip` + `latinv.zip`, tab-format:
   `<book-code>\t<chapter>\t<verse>\t<text>`), which the original (already-committed) bulk
   import must have used but which weren't preserved anywhere in-repo. Re-parsed directly from
   these archives (book-code map: `01O`-`39O` = Genesis..Malachi, `40N`-`66N` = Matthew..
   Apocalypse, `67A`/`68A`/`71A`/`73A`/`74A`/`79A`/`80A` = Baruch/Sirach/Judith/1-2
   Maccabees/Tobit/Wisdom), filling every gap the store didn't already have. 121 new chapters
   authored (118 new + the 3 hand-authored ones replaced), bringing the store to 1157 files.
   **Deliberately still left unauthored** (do not attempt to fill from this source without a
   second, independent source to cross-check against):
   - Daniel 4-14 and Esther 11-16 — this particular Vulgate digitization only has Daniel 1-3
     and Esther 1-10; the deuterocanonical continuations simply aren't in the Latin file at all
     (English/Douay-Rheims does have them).
   - 12 chapters (`bar-2`, `isa-5`, `isa-25`, `joan-15`, `rom-9`, `rom-11`,
     `sir-10`/`19`/`23`/`41`/`51`/`52`) where the Douay-Rheims side has a corrupted, misnumbered
     block — e.g. `isa-25`'s own line tagged verse 1 actually carries verse 11's text — rather
     than simply missing content. Confirmed by checking real verse content against what the
     tag claims; importing under the source's own (wrong) verse numbers would have silently
     attached mismatched text to real citations, which is worse than leaving them `missing`.
   `tests/scripture/bulk-import.test.ts`'s invariant was widened to tolerate up to 2 genuinely
   isolated missing-English verses per chapter (an existing pattern for `isa-33`/`sir-36`'s
   pointing-split half-verses, now also covering true source gaps like an in-file "Transfer
   interrupted!" download-corruption artifact found at `matt-12:43`) while still failing loudly
   on anything resembling the corrupted-block pattern above.

`npm test` (1990 passed), `tsc --noEmit`, and `npm run build` all pass.

### Bulk scripture import, take two — re-sourced the corrupted/missing chapters from drbo.org (2026-09-03, same day)

Closes the two gaps the previous entry left open. The "Unbound Bible" source used for the
first pass turned out to have real defects, not just gaps: 12 chapters where the Douay-Rheims
side had verses shifted/merged under the wrong number (confirmed by content, e.g. the line
tagged "Isaias 25:1" actually held 25:11's text), and Daniel 4-14/Esther 11-16 simply weren't
in its Latin file at all.

Found a better source: drbo.org (Douay-Rheims Bible Online) publishes a combined "DR + LV"
page per chapter (`/drl/chapter/<book><chapter>.htm`) with each verse's Latin and English kept
in the same HTML block — structurally impossible for the two languages to drift out of
alignment the way the tab-separated source could. Fetched all 28 real chapters this way
(`bar-2`; `isa-5`, `isa-25`; `joan-15`; `rom-9`, `rom-11`; `sir-10/19/23/41/51`; `dan-4`
through `dan-14`; `esth-11` through `esth-16`) via a small regex parser against the raw HTML
(`<p><a class=vn>N</a>ENGLISH<br><font class="latin">LATIN</font></p>`), spot-checked several
against the known-correct text, and confirmed `sir-52` isn't a real chapter at all (Sirach
only has 51 on drbo.org — the old source's "52" was itself bogus, dropped rather than
authored). Store now at 1185 chapter files; every book/chapter citation used anywhere in the
app resolves to real text with no remaining gaps.

`tests/scripture/scripture.test.ts`'s "unauthored chapter" case repointed from `dan-14` (now
authored) to `gen-51` (Genesis only has 50 chapters — a real "doesn't exist," same code path).

`npm test` (2018 passed), `tsc --noEmit`, and `npm run build` all pass.

### Bulk scripture import, take three — recovered the "Transfer interrupted!"-style isolated gaps too (2026-09-03, same day)

One more pass over the first-pass ("Unbound Bible" source) chapters that still had a lone
isolated missing English verse (33 chapters, 34 verses total, `matt-12:43`'s in-file "Transfer
interrupted!" artifact among them) — re-checked every one against drbo.org's per-verse text
rather than assuming they were unrecoverable. 33 of 34 turned out to be real, present verses
that the old scrape had simply dropped (a transient download/parse defect, not a source gap)
and are now filled in with verified text.

The other 10 (`2cor-1:24`, `2reg-13:39`, `2thess-2:17`, `1thess-4:18`, `gen-5:32`, `isa-46:13`,
`jdth-4:17`, `joan-11:57`, `jos-10:44`, `sir-29:35`) are confirmed genuine: drbo.org's own
Douay-Rheims numbering for these chapters stops one verse short of the Vulgate's own count too
(checked each chapter's own max verse number on drbo.org) — a real Vulgate/Douay-Rheims
versification difference, not a scrape artifact, so these stay honestly `missing`, alongside
the pre-existing `isa-33`/`sir-36` pointing-split half-verses (same convention, different
cause).

`npm test` (2018 passed), `tsc --noEmit`, and `npm run build` all pass.

### Seasonal hymn-doxology swap — audited Lauds/Vespers/Matins, wired the two real gaps (2026-09-03, same day)

Closes the "unexamined gap" logged for Lauds/Vespers/Matins: the mechanism
(`hours/hymn-doxology.ts`) was previously wired only into Prime and Compline, with every
Lauds/Vespers/Matins hymn stored as one fixed text and no check done on whether the source
actually varies its closing doxology by season there.

Audited by exhaustively grepping every `[Hymnus ...]`/`[HymnusM ...]` section in Divinum
Officium's own `Psalterium/Special/{Major,Matutinum} Special.txt` for a real "*"
substitution marker (distinct from the unrelated antiphon-incipit "*" convention used
throughout responsory/antiphon text). The Monastic 1617 track (vu's own structural source)
turned out to carry the marker in exactly two places: Sunday's plain-default Vespers hymn
("Lucis Creator", inherited unchanged from the base secular text via a `@:`-reference with
unrelated wording substitutions) and the plain ferial Matins hymn (Day1-6, all six weekday
forms share it). Every other hymn checked — Sunday Lauds, the Mon-Sat Lauds/Vespers weekday
defaults, the Advent Lauds/Vespers/Matins hymns, and (initially assumed to need this) the
Monastic Paschaltide Lauds/Vespers/Matins hymns specifically — has its own fixed,
non-swapping doxology already baked into the text in the source itself. This generalizes
Terce/Sext/None's already-confirmed "self-contained, no swap" finding without making it
universal — Lauds needed no wiring at all as a result.

Wired the two real cases via `getHymnDoxologyId(day, perAnnumId)` (signature changed from
`(season, winner, perAnnumId)` to take the whole `LiturgicalDay`, needed for the date-window
check below) into `vespers.ts`'s new `resolveVespersHymn` helper and `matins.ts`'s
`resolveMatinsHymn` fallback tier. New `vespers-hymn-doxology-per-annum.yml`/
`matins-hymn-doxology-per-annum.yml` hold each hymn's own natural per-annum ending, split out
of `vespers-hymn-sunday.yml`/`matins-hymn-ferial.yml`.

Also closed two related real gaps found during the same source audit, both live-checked
against `hymni.pl`'s `doxology()` sub (the mechanism this project's own hymn-doxology.ts is
modeled on):

- The Common of the BVM's own `Doxology=Nat` rule (`Commune/C10.txt` through `C12Q.txt`) means
  Marian Saturdays get the Nativity doxology regardless of season, not just during
  Christmastide/Epiphanytide — `hymn-doxology-by-feast.yml`'s `byFeastId` now has a
  `marian-saturday` entry, and `resolveOfficeWinner`'s feast-override check in
  `getHymnDoxologyId` now looks up a `kind: 'temporal'` winner's id too, not just
  `'sanctoral'` (a temporal id and a sanctoral id never collide, so this is safe to check
  unconditionally).
- The Assumption's octave (Aug 16-22) is hardcoded in the source to the Nativity doxology
  regardless of season or the day's own winner — confirmed *not* excluded for the 1570/1617
  track (unlike its Dec 9-15 companion rule for the Immaculate Conception's run-up, which the
  source explicitly excludes 1617 from, so that one was deliberately left unported). Checked
  directly by date (`isAssumptionOctaveWindow`), same pattern as `marian-antiphon.ts`'s
  `isCandlemasToHolyWednesday` — this window doesn't line up with any `Season` value.

`npm test` (2021 passed) and `tsc --noEmit` both pass.

### Seasonal hymn-doxology swap, take two — broadened beyond the source's own marker (2026-09-03, same day)

Follow-up to the entry above, at the user's explicit direction: "the doxology is a neat way to
stay tuned in to the season... we're doing a remix, not a copy." The first pass wired the
mechanism only where the Monastic source itself carries a real "*" substitution marker (two
spots total). This pass wires it into *every* plain-default Lauds/Vespers hymn (Sunday,
Mon-Sat, Advent, Lent, Passiontide, Paschaltide — 21 files total, all but the two already done)
regardless of whether the source swaps that specific hymn, matching this project's own
"Not a reconstruction" stance (repo `CLAUDE.md`) rather than the reference engine's literal
behavior.

Each hymn file's own natural per-annum ending was split out into a matching
`{hour}-hymn-{key}-doxology-per-annum.yml` (same convention as the first pass), so a
season/feast with no override (Advent, Lent — the source never swaps those, and nothing here
changes that) reproduces the exact original text; `lauds.ts`'s new `resolveLaudsHymn` and
`vespers.ts`'s widened `resolveVespersHymn` route every plain-default hymn through
`appendDoxology`/`getHymnDoxologyId` uniformly. The real, visible effect: hymns shown during
Christmastide/Epiphanytide/Corpus Christi/Sacred Heart (none of which `seasonalOfficeSuffix`
covers, so these were always the plain weekday-default hymns) now pick up the Nat/Epi/Corp/Heart
doxology instead of staying on their generic ending; Ascensiontide and the Pentecost octave now
correctly diverge from the rest of Paschaltide's "Pasch" doxology to their own "Asc"/"Pent" one
mid-season. Matins was left as before (only the ferial fallback wired, from the first pass) —
it has no per-weekday or per-season hymn files to broaden into; every other season there still
falls through to that same one ferial hymn already.

A first automated extraction pass had a real bug (a body/stanza-boundary parser that failed to
stop at `status:` when a hymn's English translation had one fewer stanza than its Latin — no
English doxology stanza was ever transcribed for several weekday hymns — silently swallowing
the YAML `status:` block into the hymn text) and was caught before being trusted: none of its
output was committed, all 20 affected originals were reverted via `git checkout` back to this
same day's untouched-by-that-bug state, and redone with a corrected parser plus manual
spot-checks of the regenerated files.

`npm test` (2021 passed) and `tsc --noEmit` both pass. Live-checked via a scratch test (not
committed) confirming Epiphanytide/Eastertide/Ascensiontide/Pentecost/Sacred-Heart plain days
now show the right doxology and Advent/Lent plain days are byte-identical to their pre-change
text.

### Sunday Matins Nocturn 3: bare Gospel incipits authored in full, all 48 remaining Sundays (2026-09-03)

User noticed a real gap while praying Matins on a Sunday: Nocturn 3 showed St. Augustine's
homily with no Gospel citation or text above it, just the patristic commentary starting cold.
Root cause was this project's own standing convention (see the temporal-nocturn-sweep entries
above): every Sunday's Matins source gives Nocturn 3's Gospel pericope as a bare "Et réliqua"
incipit (e.g. "Homo quidam fecit cœnam magnam... Et réliqua"), and this project's authoring
passes deliberately skipped transcribing it, keeping only the homily as an `isGospel: false`
lesson — noted at the time as "sourcing that properly would need live-CGI queries per date, out
of scope for this pass." The mechanism to *render* a paired Gospel+homily as one atomic part
already existed and worked correctly (`buildReadingPool` in `hours/matins.ts`, done 2026-08-28)
— post-pentecost-15.yml was the one file that already had a full Gospel, which is what made the
asymmetry visible as a real content gap rather than a missing feature.

Closed for all 41 remaining Sunday `nocturn-readings` files that had exactly one homily-only
Nocturn 3 entry: `post-pentecost-{02..14,16..24}` (22), `post-epiphany-{1..6}` (6),
`septuagesima`/`sexagesima`/`quinquagesima` (3), `lent-{1..4}` (4),
`easter-octave`/`easter-{3..6}` (5), `sunday-after-ascension` (1), plus Trinity Sunday
(`post-pentecost-01.yml`, which has *two* Nocturn 3 Gospel+homily pairs — its own, and the
displaced First Sunday after Pentecost's commemoration).

**Sourcing method, better than the live-CGI-per-date approach originally assumed necessary**:
the reference engine's Missale (`web/www/missa/Latin|English/Tempora/<id>.txt`) files carry the
same Sunday's full `[Evangelium]` — the Mass Gospel and the Matins Nocturn 3 Gospel are the same
pericope — keyed by the identical `<Season><N>-0` id already recorded in each `nocturn-readings`
file's own header comment (e.g. `post-pentecost-14` → `Pent14-0`). Reading that file directly is
both more reliable and far faster than per-date live-CGI queries, which turned out to be a
dead end here anyway: a live query landed on whatever Sunday the *live engine's own* calendar
assigns to that specific date, which doesn't always match vu's blended calendar's numbering for
the same nominal week (confirmed by one query returning the wrong pericope, Luke 14:26-35
instead of the expected Luke 14:16-24, for a date with a commemorated saint). One exception:
Trinity Sunday's second reading (the displaced First Sunday after Pentecost's own Gospel, Luke
6:36-42) has no live date to query at all, since Trinity permanently occupies that slot every
year — found instead in `missa/Latin/Tempora/Pent01-0a.txt`, the reference engine's own
preserved pre-Trinity-displacement Mass propers for that Sunday.

Each new entry is `isGospel: true`, inserted immediately before the existing homily entry in
the same Nocturn 3 group (satisfying `buildReadingPool`'s adjacency-pairing check with no
mechanism changes needed), citation/text sourced verbatim from the Missale bilingual pair, no
`responsory` of its own — the existing single responsory stays attributed to the homily/final
reading, unchanged, per each file's own already-documented provenance. Fixed two English-text
artifacts found while transcribing (a stray leading "v." verse-marker on three Eastertide
Gospels' English text, `easter-{4,5,6}.yml`; one line with a stray extra leading space,
`post-pentecost-19.yml`) rather than reproducing them. Every touched file's header comment got a
short "Update (2026-09)" note explaining the new sourcing, without rewriting the original
sweep-method prose.

`npm test` (2021 passed, all 47 files) and `tsc --noEmit`/`vite build` both pass. Spot-verified
via a scratch test (not committed) against `resolveOrdo('matins', ...)` on each id's real Sunday
date, confirming a single `kind: 'gospel'` part with a nested `homily` field appears (not two
split lessons) for post-pentecost-14 (the date that surfaced the bug), post-pentecost-02,
Trinity Sunday (both pairs), post-epiphany-1, septuagesima, easter-octave, and
sunday-after-ascension.

**Follow-up, same day**: a second pass over the same file shape (`nocturn: 3` / `isGospel:
false` with no paired Gospel) turned up 7 more Sundays this first pass missed —
`advent-{1..4}`, `christmas-octave-sunday`, `passion-sunday`, `palm-sunday` — bringing the
total to 48. Same method (Missale `[Evangelium]`, keyed by each file's own header-documented
id: `Adv1-0`…`Adv4-0`, `Nat1-0`, `Quad5-0`, `Quad6-0`), with one real wrinkle: **Palm Sunday's
Missale `[Evangelium]` tag is the Passion narrative** (Matt 26-27, read at Mass itself), not
the Matt 21:1-9 entry-into-Jerusalem pericope this Nocturn 3's homily (St. Ambrose on Luke,
Bk. 9) is actually about — that shorter Gospel is proclaimed earlier, at the Blessing of Palms,
embedded in `Quad6-0.txt`'s own `Benedictio Palmorum` section rather than under its
`[Evangelium]` tag. Sourced from there instead, confirmed by cross-checking the citation against
the file's own existing header comment, which already named Matt 21:1-9 as the homily's true
subject. `advent-4`'s Missale Latin text is itself a cross-reference
(`@Tempora/Adv3-6`) rather than inline — resolved by reading `Adv3-6.txt` directly, same
transparent-reference-following already used elsewhere in this project's sourcing.

Discovered one real, useful regression while verifying: `tests/calendar/
vigil-of-christmas-sunday.test.ts` asserted exactly one Matins Gospel on the rare Sundays where
Advent 4 lands on Dec 24 and the Vigil of Christmas wins outright with Advent 4 commemorated
(2023-12-24, 2028-12-24) — true only because Advent 4 previously had no Gospel content to
contribute. With Advent 4's own Gospel now authored, `buildReadingPool`'s existing (unchanged)
commemoration-pooling logic correctly surfaces *both* Gospels — Vigil's own (Matt 1:18-21, St.
Jerome) and Advent 4's own commemorated one (Luke 3:1-6, Pope St. Gregory the Great) — matching
the same generous-commemorations behavior the test already expected of Lauds's two collects on
that same date, just above it in the same file. Updated the test's assertion and its own
explanatory comment to expect and name both, rather than narrowing the new content back to fit
the old, incomplete assertion.

Full suite re-run after this follow-up: `npm test` (2021 passed, 47 files) and `npm run build`
(`tsc --noEmit` + `vite build`) both pass.

### Ember days' own Benedictus/Magnificat antiphons authored (2026-09-03)

Follow-on from the 2026-08-22 Ember-day mechanism entry, whose "not authored, logged instead"
note called the gap "the Ember-day-specific `[Ant 2]`/`[Ant 3]` Lauds antiphons" — imprecise, and
worth correcting here. Live-verified (Monastic Tridentinum 1617) against a clean year for each
Ember weekday (Wed 2023-09-20, Fri 2024-09-20, Sat 2025-09-20 for September; Dec 16/18/19 2026
for Advent, all winning outright): the day's own Lauds *psalmody* (Ps 50/63/64 + canticle +
Laudate, all 5 antiphon slots) stays entirely on the plain ferial weekday default, unchanged —
only the **Benedictus antiphon** (and, Wed/Fri only, the **Magnificat antiphon**) is genuinely
proper to the Ember day, each labeled "Antiphona ex Proprio de Tempore" live. `hours/resolve-
common.ts`'s `getBenedictusAntiphon`/`getMagnificatAntiphon` already look up `${winner.id}-
benedictus-antiphon`/`-magnificat-antiphon` unconditionally for any temporal winner (rank-
agnostic by design, same as the chapter/hymn bundle) — so this needed no mechanism change at
all, just the missing content files, which is why it silently rendered blank rather than
tripping any test.

**Authored**: 6 Benedictus antiphon files (one per Ember weekday × quarter) and 4 Magnificat
antiphon files (Wed + Fri only, both quarters). September Wednesday/Friday reuse the identical
text for both (`[Ant 3]` is a bare `@:Ant 2` cross-reference in the source, not independently
authored — a real fact about the source, not a shortcut taken here); Advent Wednesday's two
differ (`Missus est Gábriel` / `Ecce ancílla Dómini`); September Saturday uses the Monastic
track's own `TemporaM/093-6.txt` override (`Illumináre... Deus Israël`) per this project's
Monastic-1617-first preference, matching the choice already made for this day's Matins
responsories.

**Not authored, correctly**: Magnificat antiphons for either quarter's Ember Saturday — Second
Vespers on both is always ceded outright to the following Sunday ("Vespera de sequenti; nihil de
præcedenti," live-verified), and the source file itself has no `[Ant 3]` for either Saturday,
confirming the omission is real, not a gap. Advent Ember Friday's own Magnificat antiphon *was*
still authored despite being currently unreachable — Dec 18 always falls inside the O Antiphons'
own unconditional Dec 17-23 window, which wins the Vespers Magnificat outright — per this
project's standing "import available real content even if it may not survive the current
calendar" default.

New assertions added to `tests/calendar/ember-days.test.ts` (Benedictus antiphon text/status on
all 6 winning-outright dates; Magnificat status on Advent Ember Wednesday, the one case with no
overriding mechanism to complicate the assertion). `npm test` (2023 passed, 47 files) and
`tsc --noEmit` both pass.

**Follow-on, same day**: user asked why the Ember days didn't actually show anywhere in the app
on 2026-09-16/18/19 despite the above. Real bug, found by resolving `getDayLabel` directly for
those dates: `calendar/day-label.ts`'s `collectCommemorations` only ever matched a `temporal`-kind
commemoration against `resolveTemporalId(day.date)` (the plain, un-relabeled id for that date) —
never against an Ember-relabeled id, so `applyEmberDay`'s own commemoration was silently dropped
from the label every time. Worse, even the *winning-outright* case had no name: the final
fallback branch calls the generic `temporalLabel(day)` (pure season/weekday arithmetic, no id
awareness at all), so an Ember day that won outright rendered as bare "Wednesday in the Nth week
after Trinity (Feria)" with zero mention of Ember anywhere.

Fixed with a small standalone `EMBER_DAY_NAMES`/`emberDayLabel` lookup in `day-label.ts` itself
(6 entries, Latin names read straight from each day's own `[Officium]` header) — deliberately
*not* added to `temporal-feasts.ts`'s named-feast table, since that table doubles as `resolve-
common.ts`'s `isFerialOrVigil` bareness check (`isBareFeria = ... && !getTemporalFeastRecord(...)`)
and adding Ember ids there would have silently turned off their own ferial Preces (their `[Rule]`
is literally "Preces Feriales"). Wired into both places: `collectCommemorations`'s temporal
branch (demoted case) and `getDayLabel`'s final fallback (winning-outright case). New assertions
in `tests/calendar/ember-days.test.ts` cover both. `npm test` (2024 passed) and `tsc --noEmit`
both pass.

**Second follow-on, same session**: user then flagged `/vu/2026-09-18/matins` (Ember Friday,
commemorated) showing the Judith reading under a bare "Reading" heading. Real root cause, and a
much broader bug than just Judith: `data/hours/bible-plan/*.yml`'s ~390-row TSV import uses
English-style book abbreviations (`jdt`, `1kgs`, `luke`, ...) for the ~30 books where that
diverges from this store's canonical Vulgate-numbering form (`jdth`, `3reg`, `luc`, ...) —
`scripture/index.ts`'s own `BOOK_ALIASES` already normalizes this for verse-*text* lookup, but
`bible-plan.ts`'s incipit (`bible-book-incipits.ts`) and responsory (`matins-responsories.ts`)
lookups both used the raw, unaliased abbreviation, silently missing on any canonically-keyed
book and falling back to the generic "Reading" heading. The 3 Gospel aliases (`luke`/`mark`/
`john`) happened to self-heal through a second, independent citation-regex fallback that only
exists for Gospels (`matins.ts`'s `getGospelIncipitFromCitation`) — every other aliased book
(Judith, both Kings pairs, both Samuel books, Esdras, Osee, Jonas, Josue, Micheas, Abdias,
Judges, the 3 Johannine epistles, James, Jude, Philemon, Titus, Zacharias, Sophonias) had no such
fallback and was silently broken wherever it happened to be the first book of a reading. Fixed by
exporting `canonicalBook` from `scripture/index.ts` (the same substitution `getScriptureChapter`
already applies internally) and routing both `bible-plan.ts` lookups through it — one shared
normalization, not two independently-aliased lookups that can drift again. New test in
`tests/propers/bible-plan.test.ts`. `npm test` (2025 passed) and `tsc --noEmit` both pass.

Separately, the user also asked about seeing 3 lessons in a row all headed "Pope St. Gregory the
Great, Homily 33 on the Gospels, on Luke 7:36-50" (Ember Friday's own 3-lesson split of one
homily, each with its own distinct text/responsory — see the 2026-08-22 Ember-day entry above).
Confirmed this is correct content, not a duplication bug, but also confirmed it's *not* actually
how the reference source itself labels it: `Tempora/093-5.txt` only carries the attribution once,
on `[Lectio1]` — `[Lectio2]`/`[Lectio3]` have no repeated header at all, continuing silently under
the same nocturn. This app's own authoring convention repeats the full `source` string identically
on every split lesson, and does so consistently across the *entire* nocturn-readings corpus, not
just Ember days — flagged for the user as a real presentational rough edge, not fixed yet pending
their call on scope (a corpus-wide UI/authoring change, not a narrow one-file fix).

**Third follow-on, same session**: user then asked why a Sunday's own patristic homily was
showing up on a Friday at all. Pulled the live reference engine's own Matins page for St. Joseph
of Cupertino's day (Tridentine 1906) to check, rather than reasoning from vu's data alone: real
Nocturn III there is exactly 3 lessons — 2 from the winning Duplex saint's own Common homily, plus
a single "**Commemoratio Feriæ**" lesson for the displaced Ember day. No Sunday content
whatsoever. `hours/matins.ts`'s `nocturnReadingIds` had a `temporalKept` block that pooled the
governing Sunday's own id (`temporalId`) *and* its `month-week-N` sibling whenever any temporal
commemoration survived on a 3-nocturn day, with no weekday check at all — so a Duplex+ saint
winning outright on a plain weekday (not a Sunday) that also happened to carry a temporal
commemoration pulled in a whole second homily's worth of unrelated content (here: St. Ambrose's
homily + its own Gospel from `post-pentecost-16`, and St. Leo's September-fast sermon from
`month-week-093` — two different reference-engine representations of the *same* underlying
Sunday, both wrongly pooled at once). Fixed by gating `temporalKept` to `day.weekday === 'sunday'`
— the governing Sunday's own content only belongs in the pool when that Sunday is the day actually
being celebrated. Live-verified by reloading `/vu/2026-09-18/matins`: both spurious readings are
gone. `npm test` (2025 passed) and `tsc --noEmit` both pass.

Still open, by the user's own explicit request to land this first: Ember Friday's own homily
still surfaces as all 3 of its lessons even when merely commemorated, where the live engine caps
a commemorated feria at exactly one lesson (its own Lectio 1 only) — same shape of fix as this
one, not yet done.

### Ezekiel spelling fix in the book-incipit heading (2026-09-04)

User caught `bible-book-incipits.ts`'s `ezek` entry reading "A reading from the Prophecy of
**Ezechiel**" — the Vulgate/Douay-Rheims spelling, inconsistent with this app's own generated
heading text (which otherwise uses modern English book names, see that file's own header note on
"Saint Paul" vs "blessed Paul"). Changed to "Ezekiel". Left every other in-corpus "Ezechiel"
alone (scripture verse text, martyrology narrative, homily quotes under `src/data/scripture/`,
`src/data/martyrology/`, `src/data/propers/`) — those are quoted Douay-Rheims translation prose,
where the traditional spelling is correct as part of the source text, not this app's own
generated label.

### Ember September Friday's homily recombined into one reading; the Matins ferial hymn now varies by weekday (2026-09-04)

Two fixes from one report. The user looked at `093-5.txt` (Ember Friday's own source file) and
confirmed what the 2026-09-03 "second follow-on" entry above had only flagged: Gregory's Homily 33
(Lectio1-3) is one continuous, unbroken text on the same pericope (Luke 7:36-50), split only to
fill the fixed 3-lesson Nocturn slot count — exactly the "source doesn't genuinely split" case
`post-pentecost-15.yml`'s own doc comment already establishes as this store's default. Recombined
`ember-september-friday.yml`'s 3 readings into 1, keeping Lectio1's own responsory (the "first one,
when several collapse" convention) and dropping Lectio2/3's own distinct responsories, same as
every other collapsed-reading case in this store. The same "genuinely distinct proper responsory"
rationale for staying split also appears, unfixed, in `ember-september-wednesday.yml`,
`ember-september-saturday.yml`, `ember-advent-wednesday.yml`, `ember-advent-friday.yml`, and
`ember-advent-saturday.yml` — out of scope here (the user's report was specifically about
September Friday), flagged for a possible later consistency pass. Live-verified on
`/2026-09-18/matins` (Ember Friday commemorated under St. Joseph of Cupertino): one
"POPE ST. GREGORY..." heading, one flowing text, one responsory, no more triple-repeat. Updated
`tests/calendar/ember-days.test.ts`'s lesson-count assertion (now per-date, 1 for Sept 18, 3 for
the other 5 still-split Ember days).

Separately, the user reported that today's (2026-09-04, a Friday) Matins hymn started "O Three in
One, and One in Three" in the live Monastic 1617 office, not the hymn vu was showing. Investigated
and found a real, broader mechanism gap: `hours/matins.ts`'s `resolveMatinsHymn` fell through to a
single `matins-hymn-ferial` id for *every* weekday, but the source (Psalterium/Special/Matutinum
Special.txt's "Day0 Hymnus" through "Day6 Hymnus", Day0=Sunday..Day6=Saturday) genuinely varies the
plain ferial Matins hymn by weekday — the classic "week of hymns" (Somno reféctis ártubus /
Consors patérni lúminis / Rerum Creátor óptime / Nox atra rerum / Tu Trinitátis Únitas for
Mon-Fri). `matins-hymn-ferial.yml`'s own text turned out to already be Monday's hymn, mislabeled
in its header comment as a generic fixed "ferial" text and misattributed to "a plain ferial
Tuesday" (the live-verification date given, 2026-12-15, is a Tuesday, but the text itself is
Monday's — Monastic 1617's own `Day1 HymnusM` swap, "omni témpore" -> "in perpetuum", is present in
the file). Authored `matins-hymn-ferial-tuesday/-wednesday/-thursday/-friday.yml` (Monastic 1617
text, live-verified against the CGI instance directly for Friday, including its own real M-track
word-swaps: `flagitémus ómnium/vúlnerum` order and `astet`->`adsit`) and wired
`resolveMatinsHymn`'s final fallback to pick by `day.weekday`. Sunday and Saturday deliberately
left on the existing `matins-hymn-ferial` (Monday's text) as a documented placeholder, not fixed
here: Sunday's own real hymn selection is a genuinely more complex, not-yet-understood two-hymn
mechanism in the source (`Day0 Hymnus` vs `Day0 Hymnus1`, tags `H-MatDom1`/`H-MatDom4`, purpose
unconfirmed), and Saturday's own plain `Day6 Hymnus` is live-verified to be superseded on nearly
every ordinary Saturday by the Officium Sabbatorum B.M.V.'s own proper hymn ("Quem terra, pontus,
æthera") — a whole separate Saturday-votive-Office mechanism this app doesn't model at all yet.
Both are real, open gaps, not silently resolved. Updated 2 pre-existing `FERIAL_DATE` (2026-12-01,
a Tuesday) hymn assertions in `tests/hours/matins.test.ts` that had been asserting the old,
wrong-for-Tuesday "Somno" text. Live-verified `/2026-09-04/matins` in the browser. `npm test`
(2025 passed) and `tsc --noEmit` both pass.

### Matins/Lauds hymns completed for Marian Saturday; corrected 3 pre-existing content bugs found in the process (2026-09-04)

Following straight on from the "Saturday's own Officium Sabbatorum" gap flagged just above: asked
the user what the plain *secular* Tridentine 1570 track (fewer saints, easier to isolate a clean
date) does on an ordinary Saturday, expecting a straightforward answer. Found instead that a
**full precedence mechanism for this already existed**, committed well before this session
(`calendar/index.ts`'s `applyMarianSaturday`, gated on `weekday === 'saturday'`,
`temporalCategory === 'ordinary-feria'`, and no active octave; a `marian-saturday` collect and
Benedictus antiphon; a Lauds psalmody override) — a real gap in my own earlier claim to the user
that "vu doesn't model this at all," found by not having grepped the codebase before saying so.

Per direct instruction, the actual design (a real, closed decision, not left open): Marian
Saturday is rank **Simplex**, but its own separate tier — beats a plain Simplex saint (who is then
commemorated), loses cleanly (zero commemoration, matching live behavior confirmed against 3 real
dates: St. Blaise/Simplex commemorated 2024-02-03, St. Athanasius/Duplex and Ss. Cyriacus &
Co./Semiduplex both winning with *zero* trace of Marian Saturday 2026-05-02/2026-08-08) to
anything Vigil-or-higher — i.e. sits between `simplex` and `vigil` in vu's own `FeastClass` scale,
which `applyMarianSaturday`'s `winner.rank !== 'simplex'` gate already implements correctly without
needing a new `FeastClass` value at all (Vigil already ranks above Simplex there). Matins/Lauds get
proper hymns; ferial psalmody is kept everywhere (not the real historical proper-psalm set) with
only the antiphons swapped to Marian text; the Marian collect applies whenever the day wins. No
"little office" — deliberately narrower than the actual historical Officium Sabbatorum.

Authored `matins-hymn-marian-saturday.yml` ("Quem terra, pontus, ǽthera") — the one piece Matins
had nothing for at all, closing the original gap this whole investigation started from. For Lauds,
found the *hymn* ("O gloriósa Dómina") was already authored, but filed under the plain ferial
`lauds-hymn-saturday` id instead of its own `marian-saturday` id, and its capitulum/responsory/
versicle siblings were live-verified from a *different*, privileged (Lenten Ember) Saturday, not
this one — three real bugs from an earlier pass's own "believed genuine per-annum content, not
verified" comment. Fixed all three: moved the hymn to `lauds-hymn-marian-saturday.yml`, authored
real `lauds-{capitulum,responsory,versicle}-marian-saturday.yml` (live-verified 2026-10-03), and
corrected `lauds-hymn-saturday.yml`/its doxology/its capitulum/responsory/versicle siblings to the
real plain-ferial-Saturday content (live-verified 2025-09-20, a September Ember Saturday winning
outright with nothing else in play) — "Auróra jam spargit polum" hymn, Rom 13:12-13 capitulum,
"Sana ánimam meam" responsory, "Repléti sumus mane misericórdia tua" versicle; the old Isa 58:1 /
"Ipse liberávit me" / "Ángelis suis" content is now homeless, likely genuine Lenten-specific ferial
capitula this store doesn't yet model separately — not rebuilt, flagged here instead.

Per direct instruction, rolled back the pre-existing Lauds psalmody override
(`lauds-psalmody-overrides/marian-saturday.yml`) from the real historical proper psalm set (92, 99,
62 — the same numbers any duplex-majus+ feast's own Lauds override uses) to plain ferial Saturday's
own psalm groups (50, 142, the split Canticle of Moses/Deuteronomy), keeping only the antiphon
swap — matches "keep the ferial psalmody but use marian antiphons... we don't want to go
little-office" exactly. Ferial Saturday has one fewer psalm-group slot than the Common-of-BVM
pattern, so the third already-authored proper antiphon ("Nigra sum, sed formósa...") currently has
nowhere to attach and sits unused, noted in the file's own header.

Live-verified `/2026-07-04/matins` and `/2026-07-04/lauds` in the browser end to end. `npm test`
(2025 passed, one pre-existing psalm-number assertion in `tests/hours/lauds-psalmody-overrides.test.ts`
updated for the psalmody rollback) and `tsc --noEmit` both pass.

**Follow-on, same day**: user asked what actually renders on a real September Ember Saturday and on
a Saturday where a real saint outranks a merely-commemorated transferred Vigil (Aug 8 2026, Vigil
of St. Lawrence displaced by Ss. Cyriacus, Largus & Smaragdus). The second case was already correct
(the winning saint's own Common-of-Martyrs hymn applies via the pre-existing tier, unaffected by the
commemorated Vigil). The first case exposed a real gap in the weekday-hymn-cycle fix above: Saturday
was deliberately left on the `matins-hymn-ferial` (Monday's text) placeholder on the assumption it
was "almost always Marian Saturday" — true for most Saturdays, but not for a genuinely privileged
one (Ember/Advent/Lent) where nothing wins outright and Marian Saturday's own `ordinary-feria` gate
doesn't apply. Authored `matins-hymn-ferial-saturday.yml` ("Summæ Deus cleméntiæ", live-verified
against a clean September Ember Saturday, 2025-09-20) and added it to `resolveMatinsHymn`'s weekday
lookup. `npm test` (2025 passed, no test changes needed) and `tsc --noEmit` both pass.

### Sunday's own Matins hymn, plus every season's (Advent/Septuagesima/Lent/Passiontide/Paschaltide) (2026-09-04)

User asked for the actual detail behind the "genuinely more complex, not-yet-understood two-hymn
mechanism" flagged in the entry above (`Day0 Hymnus` vs `Day0 Hymnus1`). Turned out not to be a
mystery: live-checking several clean dates showed `Day0 Hymnus` ("Nocte surgéntes vigilémus omnes")
is simply Sunday's own plain per-annum hymn, and `Day0 Hymnus1` ("Primo diérum ómnium") is
Septuagesima-tide's own separate seasonal hymn — the exact same shape as Advent/Lent/Passiontide/
Paschaltide each having their own (live-verified: a Lenten Sunday and a Lenten feria both render
"Ex more docti mýstico" identically, so these seasonal hymns don't vary by weekday the way the
plain per-annum cycle does). None of the 5 seasonal Matins hymns, nor Sunday's own plain one, were
authored at all before this — every Sunday and every Advent/Lent/Passiontide/Paschaltide/
Septuagesima day (Sundays and ferias alike) that wasn't otherwise overridden by a winning saint/
octave was silently falling through to Monday's own ferial placeholder text.

Authored `matins-hymn-sunday.yml` ("Nocte surgéntes"), `matins-hymn-advent.yml` ("Verbum supérnum,
pródiens"), `matins-hymn-lent.yml` ("Ex more docti mýstico"), `matins-hymn-passiontide.yml` ("Pange,
lingua, gloriósi prœlium certáminis" — this breviary's own 5-stanza recension, not the longer
"Lustra sex..." continuation some other uses of the hymn carry), `matins-hymn-paschaltide.yml`
("Rex sempitérne Dómine"), and `matins-hymn-septuagesima.yml` ("Primo diérum ómnium") — all
live-verified, all self-contained with their own real closing doxology baked in (each one fixed and
distinct, not the shared weekday "Praesta, Pater piíssime..." ending).

Septuagesima is checked directly by `day.season === 'septuagesima'` in `resolveMatinsHymn` itself,
deliberately *not* folded into the shared `seasonalOfficeSuffix` (Lauds/Vespers' own) — that
function has no matching Septuagesima content authored for those hours yet, and `resolveOfficeBundle`
there is all-or-nothing per feast, so adding it would silently blank their chapter/hymn instead of
falling through to their current (imperfect but real) weekday default. Scoped to Matins alone; Lauds/
Vespers' own Septuagesima gap is real and separately flagged here, not fixed.

Also flagged, not fixed: the generic `matins-hymn-paschaltide.yml` will incorrectly render during
the Ascension octave specifically, which is live-verified to have its own genuinely different proper
hymn ("...Qui scandis super sídera") not yet authored anywhere — same-shape gap already existing in
Lauds' own `lauds-hymn-paschaltide.yml`, not a new regression introduced here.

Updated 2 `FERIAL_DATE` (2026-12-01, a Tuesday *in Advent*) hymn assertions in
`tests/hours/matins.test.ts` that had been asserting Tuesday's own ferial text — Advent's own
seasonal hymn correctly outranks the weekday cycle now. `npm test` (2025 passed) and `tsc --noEmit`
both pass.

### Ascension's own hymnody at Matins/Lauds/Vespers, plus the octave mechanism gap it exposed (2026-09-04)

User asked to actually close the Ascension-octave gap flagged in the entry just above. Turned out
to be a genuine mechanism gap, not just missing content: `ascension` is a *temporal*-feast id
(`data/calendar/temporal-feasts/ascension.yml`), not a saint record, so
`hours/resolve-common.ts`'s `resolveOfficeWinner` — which already synthesizes a sanctoral-shaped
winner from an active octave's own saint record (how Assumption's/St. Lawrence's octave already
gives every hour real content on interior days) — never recognized it at all. Ascension Day itself
and its whole octave were silently rendering the generic Paschaltide hymn at every hour, live-
verified against Divinum Officium (Monastic Tridentinum 1617, 2026-05-14, Ascension Day itself):
real content is "Jesu nostra redémptio" (identical at Matins/Lauds/Vespers), Act 1:1-2 (Lauds/
Vespers' shared chapter), and two genuinely distinct Lauds/Vespers responsories ("Ascéndit Deus in
jubilatióne" / "Ascéndens Christus in altum").

Fixed the mechanism two ways: (1) added `ascension` to `ALWAYS_OVERRIDE_TEMPORAL_IDS` (covers
Ascension Day itself, the same way `christ-the-king`/`circumcision`/etc. already do); (2) taught
`resolveOfficeWinner` to also synthesize a `{kind: 'temporal'}` winner from an active octave with no
saint record but a real temporal-feast record — covers ordinary-feria octave days (a real saint
still correctly keeps priority when one wins, e.g. St. Venantius/St. Peter Celestine within the
octave, live-verified unaffected). A third gap surfaced testing this: `octaveGoverningPrivilegedDay`
(the *Sunday*-side half of that same synthesis) has its own, separate, correct-by-design rule that a
privileged Sunday never accepts even an octave override — but live-verified wrong specifically for
"Dominica infra Octavam Ascensionis," which really does keep the octave's own hymn/responsory (just
with its own distinct proper chapter, not touched here). Rather than loosen that shared gate (real
risk of wrongly re-admitting some other octave's override on some other privileged Sunday, unverified
case by case), gave Lauds' and Vespers' own hymn-resolution functions (`resolveLaudsHymn`/
`resolveVespersHymn`) a narrow, hymn-only intercept: check `resolveActiveOctave` directly (no
privileged-Sunday gate) but only for a saint-less octave, before falling to the season/weekday
default — leaves every saint-based octave's existing behavior (and every other hour/part) completely
untouched.

Authored `matins-hymn-ascension.yml`, `lauds-{capitulum,responsory,hymn,versicle}-ascension.yml`,
and `vespers-{responsory,hymn,versicle}-ascension.yml` (Vespers' own chapter already falls back to
Lauds' via `vespersCapitulumForOverride`'s existing cross-hour fallback, so no separate file needed)
— all live-verified. Confirmed the whole octave end-to-end via `resolveOrdo` directly (Ascension Day,
two plain ordinary-feria octave days, the Sunday within the octave, and 3 real winning saints within
the window, all correct) rather than just the reference engine. `npm test` (2025 passed, no test
changes needed) and `tsc --noEmit` both pass.

Flagged, not fixed: Ascension's own collect is still missing (`getTemporalProper('ascension-collect')`
— a pre-existing gap, unrelated to and unaffected by this fix, not scoped to "hymnody"). Also flagged:
Pentecost/Corpus Christi/Sacred Heart are the same shape (temporal-feast octaves with no saint
record) and likely have the same latent hymn gap this fix closed for Ascension specifically — not
checked or fixed here.

**Follow-on, same day**: per direct instruction ("Ascension and Pentecost are some of the most
important material in the entire app — fix it"), closed Pentecost's own copy of the same gap
immediately, rather than leaving it flagged. Live-verified against Divinum Officium (Monastic
Tridentinum 1617), 2026-05-24 (Pentecost Sunday itself): unlike Ascension, Pentecost has 3 genuinely
different hymns, one per hour ("Jam Christus astra ascénderat" at Matins, "Beáta nobis gáudia" at
Lauds, "Veni Creátor Spíritus" — yes, that one — at Vespers), each with its own real, distinct
chapter/responsory/versicle at Lauds vs. Vespers. Added `pentecost-sunday` to
`ALWAYS_OVERRIDE_TEMPORAL_IDS` (the mechanism fix from the Ascension entry above is generic — no
code change needed beyond that one line) and authored `matins-hymn-pentecost-sunday.yml`,
`lauds-{capitulum,responsory,hymn,versicle}-pentecost-sunday.yml`, and
`vespers-{responsory,hymn,versicle}-pentecost-sunday.yml`. Confirmed the whole octave end-to-end via
`resolveOrdo` (feast day, plain octave days, and a real winning saint within the window — St.
Augustine of Canterbury, live-verified unaffected, keeps his own hymn at Lauds/Vespers while Matins
keeps Pentecost's own per the same established generous-octave design already used for Assumption/
Ascension). `npm test` (2025 passed, no test changes needed) and `tsc --noEmit` both pass.

Corpus Christi and Sacred Heart remain the one still-open piece of this same-shape gap — not fixed
this session, lower priority than Ascension/Pentecost per direct instruction.

**Follow-on, same day**: per direct instruction ("ascension's collect and the two octaves"), closed
all 3 remaining pieces immediately.

Authored `ascension-collect.yml` (live-verified 2026-05-14) — a separate, pre-existing gap
(`getDayCollect` already looked up `ascension-collect` via the same `resolveOfficeWinner` this
session's mechanism fix already covers; the collect file itself just never existed).

Closed Corpus Christi's and Sacred Heart's own hymnody gaps the same way as Ascension/Pentecost:
added `corpus-christi` and `sacred-heart` to `ALWAYS_OVERRIDE_TEMPORAL_IDS` (no further code change
needed — the mechanism is generic) and authored `matins-hymn-{corpus-christi,sacred-heart}.yml`,
`lauds-{capitulum,responsory,hymn,versicle}-corpus-christi.yml`,
`vespers-{responsory,hymn,versicle}-corpus-christi.yml`, `lauds-{capitulum,hymn,versicle}-sacred-
heart.yml`, and `vespers-{hymn,versicle}-sacred-heart.yml` — all live-verified against Divinum
Officium, 2026-06-04 (Corpus Christi) and 2026-06-12 (Sacred Heart). Corpus Christi's own hymns are
widely known outside this context too: Matins' own "Sacris solémniis", Lauds' own "Verbum supérnum
pródiens" (whose final 2 stanzas are the separately-famous "O Salutaris Hostia"), and Vespers' own
"Pange, lingua, gloriósi Córporis mystérium" (whose final 2 stanzas are the separately-famous
"Tantum Ergo").

Sacred Heart needed one real source-track correction: the Monastic Tridentinum 1617 track doesn't
observe this feast on this date at all (live-checked: shows St. Barnabas instead, a real 1617-era
calendar gap, not a vu bug) — used Divino Afflatu 1954 instead, consistent with this project's own
stated rubric/precedence track (see CLAUDE.md). Also live-verified a genuine, deliberate structural
difference: Sacred Heart's own Lauds and Vespers have **no responsory at all** in the source
(straight from chapter to hymn, "Capitulum Hymnus Versus" with no "Responsorium" between) — not an
authoring gap, so no `lauds-responsory-sacred-heart.yml`/`vespers-responsory-sacred-heart.yml` were
created. This exposed a real small UI gap: `hours/lauds.ts`'s and `hours/vespers.ts`'s own
`resolveOffice` always pushed a `{kind:'responsory'}` part even when its own resolved text carried
`status: missing` in both languages, rendering an empty section. Fixed both to omit the responsory
part entirely when both languages are genuinely missing, rather than an empty box.

Confirmed both octaves end-to-end via `resolveOrdo` (feast days, plain octave days, and real winning
saints within each window — St. Boniface/St. Barnabas within Corpus Christi's, unaffected). Found
and fixed 2 pre-existing tests whose own dates now land inside these newly-modeled octaves and were
asserting the old (buggy) fall-through behavior as if correct:
`tests/hours/lauds.test.ts`'s `FERIAL_MONDAY` (was 2026-06-08, inside Corpus Christi's octave — moved
to 2026-09-07) and `tests/hours/vespers.test.ts`'s own St. Margaret of Scotland test (2026-06-09,
also inside Corpus Christi's octave — its own title already anticipated this exact fix; updated its
assertions from the old "falls through to the plain ferial default" guess to the real, live-verified
Corpus Christi content). `npm test` (2025 passed) and `tsc --noEmit` both pass.

### Recombined the 5 remaining Ember-day nocturn-readings files (2026-09-04)

User asked "why only Wed/Sat in September?" after the earlier Ember-Friday recombine, then asked
for all the remaining smaller gaps to be closed. Checked the reference source directly for each of
`ember-september-wednesday`/`ember-september-saturday`/`ember-advent-wednesday`/`ember-advent-
friday`/`ember-advent-saturday` (all 5 still carried the original "genuinely distinct proper
responsory" split rationale) and confirmed all 5 are the same shape already fixed for Ember Friday:
one continuous, unbroken homily (Bede, Gregory, or — for the two Advent Wednesday/Friday cases —
Ambrose's Commentary on Luke, cross-referenced from the Annunciation's own Common,
`@Sancti/03-25:Lectio7/8/9`, confirmed contiguous by reading the source file directly), split only
to fill the fixed 3-lesson slot count. Recombined all 5 into single readings, same convention as
before (keep the first lesson's own responsory, drop the rest). Updated
`tests/calendar/ember-days.test.ts`'s lesson-count table (now 1 for all 6 Ember days, was a mix).
`npm test` (2025 passed) and `tsc --noEmit` both pass.

### Lent's own Lauds capitulum/responsory/versicle: not homeless after all (2026-09-04)

The "homeless Isa 58:1 / Ipse liberávit me / Ángelis suis" content flagged in an earlier entry
(believed to have no home after the Marian-Saturday/plain-ferial-Saturday content correction) turned
out to already have a real home: `lauds-capitulum-lent.yml`/`lauds-responsory-lent.yml`/
`lauds-versicle-lent.yml` already existed with this exact content, just marked `draft` (never
live-verified) rather than genuinely missing. Live-verified against both a Lenten Ember Saturday
(2026-02-28) and a plain Lenten Wednesday (2026-03-04) — identical, confirming this is genuinely
season-wide content (not weekday-varying) and correcting an earlier session's wrong "homeless" claim.
Marked all 3 files `verified`. (Vespers' own Lenten capitulum/responsory/versicle/hymn and both
hours' Lenten hymn-doxology-per-annum files are still `draft` — a separate, broader Lent-content
verification backlog noticed in passing, not fixed here, out of scope of what was specifically
flagged.)

### Trinity Sunday: a real feast, not modeled at all (2026-09-04)

A third "smaller gap" turned out much bigger on inspection: what had been described as a "Trinity
Sunday display gap" was actually a missing calendar mechanism. `getDayLabel` already had a purely
cosmetic "Trinity Sunday" label overlay for the Sunday after Pentecost, but nothing made that Sunday
actually *be* Trinity Sunday — `resolveTemporalId` just fell back to `post-pentecost-01`, an
`ordinary-sunday` category, so any Duplex+ saint assigned that date (e.g. the Queenship of the BVM,
1954, fixed May 31) won outright per the normal rule, with Trinity Sunday itself never appearing as
more than a commemoration or a cosmetic label. Live-verified against Divinum Officium (Monastic
Tridentinum 1617, 2026-05-31): the real occupant is "Festum Sanctissimæ Trinitatis, Duplex II.
classis" — one of the year's principal Sundays.

Added a real `data/calendar/temporal-feasts/trinity-sunday.yml` record (Easter+56, Duplex II.
classis). Per direct instruction, `unconditional: true` (same shape as `christ-the-king.yml`) —
Trinity Sunday always wins, since the one real collision case (the Queenship of the BVM, also Duplex
II classis, fixed May 31) isn't independently live-verifiable against the Monastic 1617 track at all
(that calendar doesn't carry the Queenship feast — May 31 there is St. Petronilla instead, confirmed
live), and a principal Sunday shouldn't lose a same-rank fixed-date collision by coincidence.

Authored the full office: `trinity-sunday-collect.yml`, `trinity-sunday-{benedictus,magnificat}-
antiphon.yml`, `matins-hymn-trinity-sunday.yml` ("Summæ Deus cleméntiæ" — shares its opening stanza
with the plain ferial Saturday hymn but diverges completely after that, a genuinely different hymn),
`lauds-{capitulum,responsory,hymn,versicle}-trinity-sunday.yml`, and `vespers-{responsory,hymn,
versicle}-trinity-sunday.yml` (Lauds/Vespers share both their chapter, via the existing cross-hour
fallback, and their hymn, "O lux beáta Trínitas" — already the pre-Urban VIII text) — all live-
verified 2026-05-31. Added `trinity-sunday` to `ALWAYS_OVERRIDE_TEMPORAL_IDS`.

Found and fixed 3 pre-existing tests written around the old cosmetic-label behavior, two of which
had never-verified guesses baked in as if correct: `tests/calendar/day-label.test.ts`'s "names the
anchor day itself" test (now Trinity Sunday wins outright over the Queenship, not the reverse) and
its own "outranks an active octave" test (previously asserted "Duplex I Class", a guess — the real,
now live-verified rank is Duplex II Class); `tests/calendar/transfer.test.ts`'s St. Felix I test
(Trinity Sunday now the real winner on 2026-05-31, Queenship/post-pentecost-01/Pentecost's octave
all commemorated in return). `npm test` (2025 passed) and `tsc --noEmit` both pass.

### Lent's remaining Vespers/Lauds draft content, verified (2026-09-04)

Closed the rest of the "Lent's own seasonal office bundle" gap flagged in the entry above the same
day: `vespers-{capitulum,responsory,hymn,versicle}-lent.yml` and both hours' own
`-doxology-per-annum` files were all real content, just never marked verified. Live-verified
Vespers' own bundle against a plain Lenten Wednesday (2026-03-04) — Joel 2:17 chapter, "Scápulis
suis obumbrábit tibi" responsory, "Audi, benígne Cónditor" hymn (including its own claimed Monastic
word-swap, confirmed live), shared "Ángelis suis" versicle. `lauds-hymn-lent.yml`'s own content
("Jam Christe sol justítiæ") and its doxology were already seen live earlier the same session (while
checking `lauds-hymn-saturday.yml`'s real content against the same 2026-02-28 Lenten Ember Saturday
date) — byte-identical, just needed the status flipped. All 8 files now `verified`; no more draft
Lent content anywhere in this app. `npm test` (2025 passed) and `tsc --noEmit` both pass.

### Paschaltide's "Suffragium Paschale" mechanism (2026-09-04)

Closed the highest-priority remaining mechanism gap. Live-verified against Divinum Officium (both
Monastic Tridentinum 1617 and Tridentine 1906, byte-identical between the two tracks — the source's
own `[Suffragium]` macro literally has a `(sed tempore paschali) @:Suffragium Paschale` override,
2026-04-20, a plain Eastertide feria): throughout the whole Paschaltide window (`eastertide`/
`ascensiontide`/`pentecost` — the same boundary `PASCHALTIDE_SEASONS` already governs elsewhere in
this store), two things happen whenever they'd otherwise show at all:

1. The ordinary 5-suffrage set (Cross/BVM/Joseph/Apostles/Peace) collapses into one single
   alleluia-form "Suffragium Paschale" block — one antiphon (different at Lauds, "Crucifíxus
   surréxit a mórtuis...", vs. Vespers, "Crucem sanctam súbiit...", unlike the ordinary 4/5
   suffrages, which share identical text at both hours), one shared versicle, one shared collect.
2. The ferial-Preces litany (`lauds-preces-feriales`/`vespers-preces-feriales`, sourced from
   Tridentine 1906/1910 since Monastic 1617 has none at all — see that file's own header) drops
   entirely rather than being replaced by anything ("Preces Feriales{omittitur}" live) — a
   Sunday/feast's own short litany (`lauds-short-litany`) is untouched, a separate, universal
   collect-opening dialogue unrelated to this ferial-specific one.

Exported a new `isPaschaltide(season)` helper from `hours/resolve-common.ts` (wraps the existing
`PASCHALTIDE_SEASONS` set, previously private to that file) and used it in both
`resolveSuffrages` (gated after the pre-existing "omit suffrages entirely" checks — Double+/active
octave/Advent-Christmastide-Passiontide already drop suffrages regardless of season, unaffected)
and a new check in `hours/lauds.ts`'s/`hours/vespers.ts`'s own `'lauds-preces'`/`'vespers-preces'`
cases (gated on `isFerialOrVigil(day)`, matching the existing predicate that already selects the
ferial litany in the first place). `resolveSuffrages` gained a third `hourId: 'lauds' | 'vespers'`
parameter (both call sites updated) to pick the right antiphon.

Authored `lauds-suffrage-paschale.yml` and `vespers-suffrage-paschale.yml` — English on the Vespers
antiphon is this project's own translation and left `draft` (the reference engine's own English
render leaves that one antiphon untranslated, Latin-only); everything else `verified`.

Verified end-to-end via `resolveOrdo` on real dates for Lauds (both the suffrage collapse and the
Preces omission fire together correctly on a clean Eastertide feria, 2026-04-20) and via a direct
`resolveSuffrages` unit call plus a separately-found clean Vespers evening (2026-04-15, since this
particular stretch of Eastertide 2026 is otherwise densely packed with First-Vespers anticipations
and octaves — Easter's own, the Patronage of St. Joseph's, Ascension's, Pentecost's — that already
correctly suppress suffrages for other, pre-existing reasons and would have made a false negative
look like a false positive). `npm test` (2025 passed, no pre-existing tests exercised this path at
all) and `tsc --noEmit` both pass.

### Matins responsory pools: per-book expansion + new per-Common-category pool (2026-09-04)

User request: real Matins responsories for each Bible book (sourced from the reference engine),
and for the patristic (Nocturn 2/3 homiletic) readings too. Investigation found the patristic side
was a much bigger, previously-untracked gap than the per-book pool's own "expand beyond `isa`" item
already on this list: 272 of 355 `data/propers/nocturn-readings/*.yml` files have no `responsory`
field at all — most of the calendar, not a handful of stragglers.

**Per-book pool** (`data/hours/matins-responsories-by-book.yml`): added `gen` (2 entries, Latin/
Tempora/Quadp1-1.txt — Monday of Septuagesima week, the historical start of the Genesis "Historia"
cycle, a plain ferial with no overriding feast; only 2 entries because the source's own English
translation is missing a 3rd) and `jer` (3 entries, Latin/Tempora/Quad6-2.txt — Tuesday of Holy
Week, "Feria privilegiata," genuine ferial Jeremiah/Lamentations content). Deliberately did *not*
seed `jer` from Pent02-5.txt despite it citing Jeremiah too — that day's Rank is Sacred Heart
(Duplex I classis), so its Responsory1-3 are the feast's own proper text, not genuine ferial-book
content, even though its Lectio verses are borrowed from Jeremiah. Still only 3 of 72 books used
in the user's own bible-plan seeded (`isa`, `gen`, `jer`) — the rest is deferred bulk content, one
real live-sourced query per book, same pattern.

**New per-Common-category pool** (`data/hours/matins-responsories-by-common.yml` +
`getResponsoryForCommon` in `src/propers/matins-responsories.ts`, sibling to the existing
`getResponsoryForBook`): keyed by `SaintRecord.common` instead of a scripture book, same "loose
match, cycle by index" design. Unlike the per-book pool (scattered one query per book, across the
whole liturgical year), a Commune category's responsories live in one static file each
(`divinum-officium-reference`'s `web/www/horas/{Latin,English}/Commune/CN.txt`) — 9 files
(`C1`-`C8`, `C11`) transcribed directly, cover all 19 of vu's real `common:` category ids (several
are genuine aliases, confirmed by reading each candidate file's own `@Commune/CX` inheritance line
rather than guessed — e.g. `common-of-a-confessor-doctor`/`-pope` inherit C4 unchanged, same pool
as `common-of-a-confessor-bishop`; `common-of-an-abbot`/`-not-bishop`/`several-confessors` inherit
C5 unchanged; `common-of-a-martyr` inherits C2 unchanged; `common-of-a-virgin` inherits C6
unchanged; `common-of-a-widow`/`several-women-martyrs` inherit C7 unchanged — see the data file's
own header for the full mapping). Not seeded: `common-of-an-angel`/`-an-evangelist`/`-a-vigil` and
the `proper-to-*` ids — none of these are real Commune categories with a responsory file to source
from (angels/evangelists have their own proper content instead).

Wired into `hours/matins.ts`'s `nocturnReadingPart`/`gospelReadingPart`: when a reading's own
`responsory` field is absent, falls back to `getResponsoryForCommon(getSaintRecord(id)?.common,
index)` — `index` is the reading's position within that same id's own file (a new `commonPoolIndex`
counter in `buildReadingPool`'s assembly loop), so a saint with several undated readings still gets
distinct responsories across them rather than the pool's first entry repeated. A temporal/octave id
(no `common` category) still correctly gets `undefined` — an honest absence, not a wrong guess.

Live-verified end-to-end via `resolveOrdo`: St. Francis de Sales (2026-01-29, `common-of-a-
confessor-bishop`, previously responsory-less) now resolves a real Nocturn 3 responsory; Ss.
Cornelius and Cyprian plus Ss. Euphemia, Lucy, and Geminianus (both on 2026-09-16, both `common-
of-several-martyrs`) each resolve their own responsory from the same pool, confirming the
per-saint-file cycling index doesn't leak across different saints sharing a day. `npm test` (2025
passed) and `tsc --noEmit` both pass.

### Matins responsory pools: 8 more books seeded (2026-09-04)

Follow-up to the same-day entry above, continuing the per-book pool's bulk-content pass — same
"one real source per book" method, no mechanism changes. Added `exod` (Latin/Tempora/Quad4-0.txt,
Sunday IV in Lent, its own ordinary rank — the burning bush through the Song of the Sea), `prov`
(081-1.txt, August week 1's own ferial), `tob` (093-0.txt, September week 3's own Sunday), `ezek`
(111-1.txt, November week 1's own ferial), and `1macc` (101-1.txt, October week 1's own ferial) —
all bilingual, all genuinely their own book's content, no overriding rank.

Also added `sir` and `sap` (identical to `prov`'s own entries) and `dan` (identical to `ezek`'s) —
not guessed reuse, confirmed by reading each week's own reference-engine file: the August "Wisdom
literature" weeks (081 Prov, 083 Sap, 084-085 Sir) and the November "prophetic" weeks (111 Ezek,
113 Dan) `@`-reference *the same* underlying responsory set regardless of which specific book that
week's Lectio actually reads that day — a real confirmation, from the reference engine's own data,
that this app's "loose, book-level match, not exact-citation" responsory design (see the pool's own
header, and the 2026-08 user direction it's sourced from) matches how these responsories genuinely
worked historically, not an invention to paper over a smaller pool.

Skipped, deliberately: `1sam` (1 Reg) — every Pent01 week file's responsories are macro references
(`@:Responsory1_ (sed rubrica cisterciensis) ...`, `[Rule] Initia cum responsory`) rather than
literal text, needing the base macro template resolved with per-file substitutions to get real
text; not worth the complexity for one book this pass. `2sam`/`1kgs`/`2kgs`/`2macc` (2/3/4 Reg, 2
Mac) — real Latin responsories exist in every week's own file, but the reference engine's own
English translation is entirely absent for all of them (checked every Post-Pentecost week each book
appears in); left unseeded rather than guessing a translation, consistent with only seeding books
that have real bilingual source text. `zach`/`amos`/`agg` (minor prophets) — same English-translation
gap, but Latin is real and available (Latin/Tempora/114-4.txt, 115-4.txt, 115-5.txt) if a future
pass wants to add an own-translation draft English side (see `vu-prefer-text-over-missing` memory
for the precedent).

Now 11 of 72 bible-plan books seeded: `isa`, `gen`, `jer`, `exod`, `prov`, `sir`, `sap`, `tob`,
`ezek`, `dan`, `1macc`. `npm test` (2025 passed) and `tsc --noEmit` both pass; live-verified via
`getResponsoryForBook` resolving real text for every new key.

### Matins responsory pools: a real extraction bug found, 8 more books unblocked (2026-09-04)

User pushback on the previous entry's "5 books skipped, no English available" claim ("look at the
secular office for most of the books") — right call: re-checking found a genuine bug, not a real
source gap. The extraction script's Responsory-block regex only matched a bare `[ResponsoryN]` tag;
it missed the reference engine's own `[ResponsoryN_]` convention (an underscore-suffixed literal
template tag holding the real text, which a plain `[ResponsoryN]` often just `@:`-references
alongside a Cistercian-rubric variant) — so every file using that convention looked like it had no
English translation when the real text was sitting right there under a different tag name. Fixed
the script to prefer the `_`-suffixed tag's literal body over a bare `@:`-reference.

This unblocked `1sam` (Latin/Tempora/Pent01-1.txt, its own genuine 3-entry set — the young Samuel's
call and consecration) outright. Chasing the `@Tempora/X:ResponsoryN` cross-file references (not
resolved by the extraction script, hand-followed instead) also unblocked:
- `2sam`/`1kgs`/`2kgs` (2/3/4 Reg) — all three books' own week files `@`-reference the *same*
  single entry, Tempora/Pent03-1Feria's Responsory1 (2 Samuel 24's census/plague narrative,
  "Recordáre, Dómine, testaménti tui") — read directly, not guessed, same "loose match" reuse
  pattern already established for the Wisdom-literature and Ezekiel/Daniel weeks. One entry only;
  Responsory2/3 chain further references not followed this pass.
- `2macc` — reuses `1macc`'s own 3-entry set outright (Tempora/104-1 `@`-references Tempora/101-1).
- `zach` (Tempora/115-5, itself referencing 111-0's Responsory4-6 — Ezekiel's own Sunday, distinct
  from `ezek`/`dan`'s own Monday-sourced entries) and `amos`/`agg` (both referencing Tempora/
  Pent01-0's Responsory1 — Isaiah's throne-vision — plus 111-0's Responsory2-3) — all three
  minor-prophet books previously assumed to have no reference-engine English at all; that was true
  of their *own* week files in isolation, but false once their `@`-references were actually
  followed.

19 of 72 bible-plan books now seeded (added this pass: `1sam`, `2sam`, `1kgs`, `2kgs`, `2macc`,
`zach`, `amos`, `agg`). `npm test` (2025 passed) and `tsc --noEmit` both pass; live-verified via
`getResponsoryForBook` resolving real text for every new key, at every pool index (not just index
0) for the 3-entry sets.

Still genuinely unseeded after this pass — real gaps, re-confirmed: books with no reference-engine
citation anywhere in Tempora at all (Josue, Judices, Ruth, Paralipomenon/Chronicles, Esdras/
Nehemias, Canticles — the historic secular/monastic Historia cycle apparently never assigns these
their own continuous reading), and the remaining ~45 New Testament epistle/Acts/Apocalypse books
plus minor books not yet attempted.

### Matins responsory pools: 12 more OT books, and 2 real process misses caught (2026-09-04)

User caught two things this pass, neither of them a source gap:

1. Directly pointed at `web/www/horas/English/Tempora/094-3.txt` and asked why Judith ("jdt") was
   listed as unseeded when it plainly has real responsory content. Checking confirmed it — worse,
   `jdt` and `esth` (Esther) had *both* already surfaced in this same day's own full-citation dump
   of Tempora (`!Jdt`, `!Esth` were right there in the list) and were simply never followed up on.
   Not a search-methodology gap this time, an attention gap: the data was already in hand.
2. Self-audit prompted by that catch found a second, independent miss: `job` — one of the
   heaviest-used books in the bible-plan (42 rows) — had been extracted early in this same day's
   session (Latin/Tempora/091-1.txt, the classic "Job Responsories," also used in the Requiem) but
   never actually committed to the pool file at all. Sourced, verified, then silently dropped.

Fixed both, then used the same "reused across a whole prophetic-literature week" pattern already
established for `ezek`/`dan`/`zach` to sweep the rest of November's minor prophets efficiently by
following each book's own `@Tempora/X:ResponsoryN` reference rather than re-querying from scratch:
`obad` (reuses `zach`'s own R1), `jonah`/`mal` (share one entry, Tempora/111-0:R7), `hos`/`nah`
(reuse `ezek`/`dan`'s own R1), `mic` (reuses `amos`/`agg`'s own R1), `hab`/`joel` (share one entry,
Tempora/111-2:R1), `zeph` (Tempora/111-3:R1, its own). Also added `eccl` (Tempora/081-2:R1, the
same August "Wisdom literature" week cluster as `prov`/`sir`/`sap`, a different entry than
`prov`'s own) and `lam` (Latin/Tempora/Quad6-4.txt — Maundy Thursday's own Tenebrae Matins, the
classic "In monte Oliveti"/"Tristis est anima mea"/"Ecce vidimus eum" responsories; unlike the
Sacred Heart/Jeremiah override case above, Holy Week's own high rank isn't an unrelated feast
displacing ordinary content here — Tenebrae's Lamentations reading is intrinsically Holy Week's
own, so no override caution applied).

33 of 72 bible-plan books now seeded (added this pass: `jdt`, `esth`, `job`, `eccl`, `hab`, `joel`,
`zeph`, `obad`, `jonah`, `mal`, `hos`, `mic`, `nah`, `lam` — 14 new). `npm test` (2025 passed) and
`tsc --noEmit` both pass; live-verified via `getResponsoryForBook` resolving real text for every
new key.

Confirmed still genuinely absent (re-checked directly, not just re-asserted): `bar` (Baruch),
Josue, Judices, Ruth, Paralipomenon/Chronicles, Esdras/Nehemias, Canticles — no citation anywhere
in Tempora. 39 remaining books not yet attempted: Leviticus/Numbers/Deuteronomy, the four Gospels,
Acts, Apocalypse, and the NT epistles.

### Matins responsory pools: real key-alias bug found and fixed (2026-09-04)

A serious correctness bug, found by re-verifying through the *real* pipeline rather than trusting
the earlier direct-function-call "live verification": `bible-plan.ts` normalizes a reading's book
through `canonicalBook()` (`scripture/index.ts`'s `BOOK_ALIASES` map) *before* calling
`getResponsoryForBook` — so the pool must be keyed by the canonical form, not the bible-plan TSV's
own raw book code, whenever the two differ. 12 of this same day's 33 seeded keys used the raw form
instead: `1macc`/`2macc` (canonical `1mach`/`2mach`), `1sam`/`2sam`/`1kgs`/`2kgs` (`1reg`/`2reg`/
`3reg`/`4reg`), `jdt` (`jdth`), `zeph` (`soph`), `obad` (`abd`), `jonah` (`jon`), `hos` (`osee`),
`mic` (`mich`) — every one of these would have silently resolved to `undefined` in the real app
despite passing this same day's own "live-verified" test, because that test called
`getResponsoryForBook` directly with the raw code, bypassing the normalization the real call site
always applies. Renamed all 12 keys (and their own header comments) to the canonical form.
Re-verified this time through the actual `getBiblePlanReadings` call site (not a direct
`getResponsoryForBook` call) on a real date (2026-06-04, `post-pentecost-01`/Thursday, `1sam 9-10`)
— confirms a real responsory now resolves where it silently wouldn't have before this fix. `npm
test` (2025 passed) and `tsc --noEmit` both pass.

**Lesson for future sourcing passes**: when a book has a `BOOK_ALIASES` entry, the pool key is the
*alias target*, not the bible-plan YAML's own book code — check `scripture/index.ts`'s
`BOOK_ALIASES` map before choosing a key, and verify through `getBiblePlanReadings`/`resolveOrdo`
(the real call chain), not a direct, unnormalized call to `getResponsoryForBook`.

### Matins responsory pools: NT books (Acts, Apocalypse, James, Hebrews, the 4 Gospels) (2026-09-04)

Continuing the same bulk-sourcing pass, applying the previous entry's own lesson from the start
this time (canonical keys, pipeline-verified). Surveyed the whole reference engine's own citation
set precisely (an `awk` one-liner reading the line *after* every `[LectioN]` tag, following two
lines deep for Gospel entries whose citation sits below a "Léctio sancti Evangélii secúndum..."
line) rather than grepping loosely for a book prefix — the loose approach had already produced one
false positive this session (`Rom` matching a Lauds Capitulum, not a real Matins Lectio).

Added, all bilingual, all real Nocturn responsories:
- `act` (Ascension Day's own Matins, Latin/Tempora/Pasc5-4.txt) and `marc` (Easter Sunday's own
  Matins, Pasc0-0.txt) — both high-ranked feast days, but genuinely proper content (Ascension's
  own Nocturn 1 *is* Acts 1; Easter's *is* Mark 16), not an unrelated override like the Sacred
  Heart/Jeremiah case — same reasoning already used for `lam`.
- `heb` (Good Friday's own Tenebrae Matins, Quad6-5.txt) — same reasoning again, Passion-narrative
  responsories ("Omnes amici mei," "Velum templi scissum est").
- `apoc` (Easter III Sunday, its own ordinary rank, Pasc3-0.txt) and `jas` (Easter IV Sunday,
  Pasc4-0.txt) — genuine Paschaltide Sunday responsory sets, not thematically tied to Apocalypse/
  James specifically (a generic Alleluia-form psalm-based set), same "loose match" convention as
  everywhere else in this pool.
- `luc`/`matt`/`joan` (Advent I/II/III Sundays, Adv1-0/Adv2-0/Adv3-0.txt) — each Gospel's own
  reading week, generic Advent-themed responsories, only 1 entry each (a second Sunday's own set
  wasn't chased down this pass).

41 of 72 bible-plan books now seeded. `npm test` (2025 passed) and `tsc --noEmit` both pass;
live-verified through the real `getBiblePlanReadings` call chain this time (not a direct
`getResponsoryForBook` call) — `easter-3`/Monday (`act`, `joan` via the `john`→`joan` alias),
`post-pentecost-01`/Saturday (`1sam` via alias, `prov`, `matt`), and a calendar-dated Jan 10 row
(`heb`) all resolve real responsory text; `Deut` correctly still resolves `undefined` (genuinely
unseeded, confirmed absent from Tempora entirely).

Confirmed no English exists anywhere in the reference engine for Galatians/Ephesians/Philippians/
Colossians/Titus/Philemon (the whole post-Epiphany reading season) or Romans (the Christmas-Octave
season) — real Latin text exists for all of these, just no bilingual source; left unseeded rather
than own-translating six-plus books in one pass. 32 books remain (9 of them the already-confirmed-
genuinely-absent `bar`/Josue/Judices/Ruth/Chronicles/Esdras-Nehemias/Canticles): Leviticus/
Numbers/Deuteronomy, Romans, 1-2 Corinthians, Galatians, Ephesians, Philippians, Colossians, 1-2
Thessalonians, 1-2 Timothy, Titus, Philemon, 1-2 Peter, 1-3 John, Jude not yet attempted.

### Matins responsory pools: last real batch (1 Cor, 2 Cor, 1 Peter, 1 John), a repeated key bug (2026-09-04)

Confirmed Leviticus/Numbers/Deuteronomy genuinely absent from Tempora (re-checked precisely with
the `awk`-based Lectio-citation survey, not a loose grep — no hit under any Latin form of the
book's name). Added `1cor` (Maundy Thursday's own Nocturn 3, Latin/Tempora/Quad6-4.txt
Responsory7-9 — the Institution narrative, distinct from that same file's Responsory1-3 already
used for `lam`), `2cor` (Epiphany II Sunday, its own ordinary rank, Epi2-0.txt), `1pet` (1 Peter's
own reading week `@`-references the ordinary Paschaltide Sunday psalm-set already used for `jac`
below — R4-6 of that same Pasc4-0.txt), and `1joan` (Sunday within the Ascension Octave, Pasc6-0.txt,
the Paraclete responsory, only 1 entry).

**A repeated mistake, caught by self-audit this time**: the previous entry's `jas` key was wrong —
same bug as the 12-key fix two entries up. `BOOK_ALIASES` maps bible-plan's `jas` to canonical
`jac`; the pool needed `jac:`, not `jas:`. Fixed, and re-verified through `getBiblePlanReadings`
(not a direct call) on `easter-6`/Monday, 2026-05-11 (`Jas 2` now resolves a real responsory).
**Ran a full audit this time**: checked every single current pool key against `BOOK_ALIASES`
programmatically (not just the ones touched this pass) — confirmed clean, zero raw-alias-source
keys remain anywhere in the file.

45 of 72 bible-plan books now seeded. `npm test` (2025 passed) and `tsc --noEmit` both pass.

Remaining 27, none easily gettable this pass: 12 confirmed genuinely absent from the reference
source (`bar`, Canticles, Joshua, Judges, Ruth, 1-2 Chronicles, Ezra, Nehemiah, Leviticus, Numbers,
Deuteronomy); 15 with real Latin but no reference-engine English anywhere (Romans, Galatians,
Ephesians, Philippians, Colossians, Titus, Philemon, 2 Thessalonians, 2 Timothy, 2 Peter, 2-3 John,
Jude). Own-translating any of these would be the next step, if wanted — not attempted here per the
established "only seed real bilingual source" pattern.

### Matins responsory pools: the "untranslated 15" turned out mostly already bilingual (2026-09-04)

Asked to translate the 15 "real Latin, no English" books. Before translating anything, re-checked
each one's own `@Tempora/X:ResponsoryN` reference chain (the same technique that unblocked 8 books
earlier the same day) — and found the whole post-Epiphany "ordinary time" reading season (weeks
3-5, covering Galatians/Ephesians/Philippians/Colossians/1-2 Timothy/Titus/Philemon/2
Thessalonians) doesn't have its own distinct responsory content at all: every one of its 21 feria
files `@`-references 2 Corinthians' own week (Epi2-0 through Epi2-6, already fully bilingual from
earlier this same day) by weekday position. Confirmed by reading every week-3/4/5 feria file's own
`@`-line directly, not guessed. Same story for 2 Peter, 2-3 John, and Jude — all four `@`-reference
Ascension Day's own responsories (already seeded as `act`). **None of these 12 books needed any
translation at all** — added `gal`/`phil`/`1tim` (share Epi2-1), `col`/`2tim` (share Epi2-2),
`eph` (Epi2-3), `1thess`/`tit` (share Epi2-4), `2thess`/`phlm` (share Epi2-6), `2pet`/`2joan`/
`3joan` (share `act`'s R1), `jud` (`act`'s R7) — all real, already-transcribed bilingual text.

Also fixed a real, separate bug the user caught by eye in the rendered app: `getResponsoryForBook`'s
`joinResponsory` concatenated the pool's `r`/`v` text with a bare newline and no "R."/"V." labels
at all — every other responsory source in the app (`octave-readings.ts`'s own `responsory` field,
every `nocturn-readings/*.yml`'s inline `responsory`, and the new per-Common-category pool) uses
the full "R. ...\n* ...\nV. ...\nR. ..." convention. Fixed by computing the repeat line from the
`*` marker already present in the pool's own `r` text (a separate commit, `1867f73`, before this
content pass).

**Only `rom` (Romans) turned out to have no content to translate at all**, on inspection — re-
checked every one of its 15 citing files (all in the Christmas-Octave `Nat0X` range): every single
one `@`-references an unrelated feast's own proper responsory (Circumcision, Christmas Day, or a
January 7 saint), never anything Romans-specific, standalone or borrowed. There's no genuine
"Romans content" anywhere in this reference source to translate, sourced or original — using one of
these borrowed responsories would repeat the Sacred Heart/Jeremiah mismatch this file already
avoids elsewhere. Left `rom` unseeded; it would need an entirely original composition, not a
translation, if ever wanted.

59 of 72 bible-plan books now seeded — every book with any genuinely usable Latin at all. The
remaining 13 (`rom` plus the 12 already-confirmed-absent books) are the real, final gap: `rom` has
no book-specific content anywhere in this source; the other 12 have no citation anywhere at all.
Ran the full canonical-key audit again (programmatic check against `BOOK_ALIASES`, per the
established lesson) — confirmed clean. `npm test` (2025 passed) and `tsc --noEmit` both pass;
live-verified through `getBiblePlanReadings` (`post-epiphany-3`/Wednesday `Gal`, `post-epiphany-5`/
Monday `1tim`, `sunday-after-ascension`/Saturday `Jude` all resolve real responsory text; `Ruth`
correctly still resolves `undefined`).

### Matins responsory pools: `cant` (Canticles) found in the Sancti files, not Tempora (2026-09-05)

User pushback on `cant` being listed "confirmed genuinely absent": right instinct to check the
Commune files specifically — Canticles/Song of Songs imagery is heavily used in Marian propers.
Checked `Latin/Commune/C12.txt` (the Little Office of the BVM) directly: it does cite `!Cant 6:8`,
but only as a Lauds Capitulum, not a Matins Nocturn lesson — Capitula don't carry their own
9-Responsory structure, and C12's actual Lectio1-3 are Sirach 24, not Canticles (matching this
pool's existing `common-of-the-bvm` sourcing). So the Commune search itself came up empty for a
genuine Matins responsory — but it redirected the search productively: checked `Sancti/` (individual
saint/feast files) with the same `awk` Lectio-citation survey used for Tempora, and found real
`!Cant` Nocturn lessons on the Visitation (July 2) and throughout the Assumption Octave
(Sancti/08-15t through 08-22t) — the historic BVM feasts that read straight through the Song of
Songs. Only the Visitation's own file (`07-02.txt`) has both languages; the Assumption Octave days
have real Latin but no English anywhere.

Added `cant`, sourced from `Sancti/07-02.txt` — genuinely Canticles-sourced responsories that
directly blend Song of Songs verses with the Visitation narrative itself ("Surge, própera, amíca
mea... Intrávit María in domum Zacharíæ"), not a borrowed-unrelated-feast mismatch.

60 of 72 bible-plan books now seeded. `npm test` (2025 passed) and `tsc --noEmit` both pass;
live-verified through `getBiblePlanReadings` (a Dec 27 calendar-dated row resolves real `Cant 3`
responsory text). Canonical-key audit re-run, clean (`cant` has no `BOOK_ALIASES` entry).

Prompted by finding `cant` in Sancti, re-checked the other 11 already-confirmed-absent books
against Sancti too (not just Tempora) — found 2 more: `lev` (Leviticus, `Sancti/02-02.txt`, the
Purification of the BVM — its own Leviticus purification-law reading is the feast's actual reason
for being) and `num` (Numbers, `Sancti/09-14.txt`, the Exaltation of the Holy Cross — the
bronze-serpent/Cross typology, Num 21:6-9, only 1 entry). Both genuine thematic matches, not
borrowed-unrelated content. The other 9 (`bar`, Josue, Judices, Ruth, 1-2 Chronicles, Ezra,
Nehemiah, Deuteronomy) checked against the reference engine's own full Sancti citation list —
confirmed still absent, no alternate abbreviation missed.

62 of 72 bible-plan books now seeded — the real final count. `npm test` (2025 passed) and
`tsc --noEmit` both pass; live-verified through `getBiblePlanReadings` (`lent-3`/Sunday resolves
real `Lev 1` responsory text, `palm-sunday`/Sunday resolves real `Num 25-27` text). Canonical-key
audit re-run, clean. Remaining 10, the true final gap: `rom` plus 9 books with no citation
anywhere in Tempora or Sancti (`bar`, Josue, Judices, Ruth, 1-2 Chronicles, Ezra, Nehemiah,
Deuteronomy).

### Advent reading-plan carryover into Christmastide (2026-09-05)

User-flagged wrinkle in `data/hours/bible-plan/*.yml`'s own Isaiah plan: it's authored as a full
4-week Advent (`advent-1-sunday` through `advent-4-saturday`, 28 rows), but real Advent is never
that long. Advent's last real day is always Dec 23 (Dec 24 is the Vigil of Christmas's own
temporal id, `calendar/temporal-id.ts`), while Advent 1 Sunday itself — the Sunday nearest Nov 30
— falls anywhere from Nov 27 to Dec 3. Even the *longest* possible Advent (Nov 27, 27 real days)
is one short of the full 28-row plan; the *shortest* (Dec 3, 21 real days) is a whole week short.
Every single year, at least the plan's last row never lands on its own natural date — not just
"most years," as the request first put it.

Fixed in `propers/bible-plan.ts`'s new `adventCarryoverReadings`: whatever doesn't land on its own
natural date now carries forward one reading per day starting Christmas Day itself, pooled
alongside whatever Christmastide's own calendar-dated rows already have for that day (never
replacing them) — same "every contributing source pools, none override" design already used
throughout this file and `hours/matins.ts`'s own reading-pool assembly. Computed live from
`calendar/temporal.ts`'s `adventStart(year)` for each date's own year (the one place in this file
that needs real calendar-year awareness, since every other lookup here is year-independent); no
new content authoring needed, since the Isaiah rows were already there, just never reachable on
their own natural date in a short year.

User's own framing accepted as the trade-off: "for advent only, we need to show all of those
chapters" — i.e. every planned reading should surface somewhere, even if that means spilling into
Christmastide, rather than silently dropping the tail end of a short year's Advent plan.

Live-verified across the full spread of possible Advent-1-Sunday dates (`adventStart` for
2020-2035 spans exactly the Nov 27 - Dec 3 range): 2028 (Advent 1 Sunday = Dec 3, the latest/
shortest case) carries all 7 of week 4's Isaiah readings forward across Dec 25-31 in order; 2027/
2032 (Advent 1 Sunday = Nov 28, near-earliest/longest) carry only the last 2. Added 4 permanent
tests to `tests/propers/bible-plan.test.ts` covering both extremes, the pooling-not-replacing
behavior, and that nothing carries over before Christmas Day itself. `npm test` (2029 passed,
+4 new) and `tsc --noEmit` both pass.

### Matins responsory pools: `deut` — the first genuine original composition, not a transcription (2026-09-05)

User found a real lead while reading a patristic homily (`English/Sancti/01-02.txt`, on Matt
23:37 "O Jerusalem, Jerusalem"): it quotes Deut 32:11's eagle image in passing ("as also it is
written... an eagle stirreth up her nest... spreadeth abroad her wings"). Confirmed the verse
number against `data/scripture/deut-32.yml`'s own `n: 11` — genuinely Deut 32:11 in standard
citation, not the breviary's own internal chant-verse-numbering for the same text (which splits
it as 32:15-16 in `Latin/Psalterium/Psalmorum/Psalm226.txt`, the real Canticle of Moses sung at
Lauds — that file supplied real, properly accented liturgical Latin and its English counterpart
gave the matching translation, so no accentuation had to be guessed).

Unlike every other entry in this pool, Deuteronomy genuinely has no historical Matins responsory
anywhere in the reference source (confirmed absent from both Tempora and Sancti in the two prior
entries) — so this one is composed, not transcribed: a real R./V. pairing built entirely from
that same canticle's own genuine verses (32:15's eagle image as the R., 32:13-14's "apple of his
eye" as the V.), verbatim text, not paraphrased or invented, just newly paired. Documented plainly
in the data file's own header comment as vu's own composition rather than presented as a
historically-attested responsory, matching this project's "not a reconstruction" ethos.

63 of 72 bible-plan books now seeded (`deut` added). `npm test` (2029 passed) and `tsc --noEmit`
both pass; live-verified through `getBiblePlanReadings` (`easter-3`/Saturday resolves the real,
properly R./V.-formatted text for `Deut 33-34`). Canonical-key audit re-run, clean (`deut` has no
`BOOK_ALIASES` entry).

### Matins responsory pools: `rom` — the second (and last) original composition (2026-09-05)

Asked directly for thoughts on Romans, the one book with real citations but no usable Matins
content anywhere (every citing file borrows an unrelated feast's own proper responsory — Christmas
Day, Circumcision, or a January 7 saint). Same approach as `deut`: composed rather than
transcribed, but built from real, already-liturgical text rather than invented — Rom 13:11-12
("the night is past, the day is at hand... let us cast off the works of darkness, and put on the
armour of light") is the standing year-round ferial Lauds Capitulum
(`Latin/Psalterium/Special/Major Special.txt`'s own "Feria Laudes") and is used again at Advent
I's own Capitulum (`Tempora/Adv1-0.txt`) — genuine accented Vulgate/Douay-Rheims text, just never
combined into its own 9-lesson Matins responsory set the way every other book in this pool was.
User-confirmed before adding.

**64 of 72 bible-plan books now seeded — every book actually used in the user's own bible-plan has
a real responsory, whether transcribed (62) or composed from genuine liturgical text (`deut`,
`rom`).** The remaining 8 are the true, final, unclosable gap: books with no citation anywhere in
Tempora or Sancti at all (`bar`, Josue, Judices, Ruth, 1-2 Chronicles, Ezra, Nehemiah) — there's no
comparable "already-liturgical, just never assembled into a responsory" text to build from for any
of these, since they're never quoted or read in the Office at all, historically. `npm test` (2029
passed) and `tsc --noEmit` both pass; live-verified through `getBiblePlanReadings` (a Jan 1
calendar-dated row resolves real, correctly R./V.-formatted text for `Rom 7-8`). Canonical-key
audit re-run, clean.

### Matins responsory pools: `bar` found in the Missa, not the Horas (2026-09-05)

The "closes the pool" claim above turned out premature by one book. User pointed to
`missa/Latin/Tempora/Quad6-6.txt` — the *Mass* propers, a directory tree this whole responsory
search had never once checked (everything so far searched `horas/`, the Divine Office proper).
The Easter Vigil Mass's own fourth prophecy (`[Proph_Baruch3]`, Bar 3:9-38, "Audi, Israël, mandáta
vitæ") is genuine, real, accented liturgical Latin — just never part of the Office's own Matins at
all, in either directory tree. Composed `bar` the same way as `deut`/`rom`: built from that
prophecy's own climactic close ("Hic est Deus noster... Post hæc in terris visus est, et cum
homínibus conversátus est" — read at the Vigil because early exegetes saw "seen upon earth, and
conversed with men" as a prophecy of the Incarnation), genuine text, not invented.

**65 of 72 bible-plan books now seeded.** `npm test` (2029 passed) and `tsc --noEmit` both pass;
live-verified through `getBiblePlanReadings` (`palm-sunday`/Thursday resolves real, correctly
R./V.-formatted text for `Bar 1-2`). Canonical-key audit re-run, clean (`bar` has no `BOOK_ALIASES`
entry).

**This closes the per-book responsory pool as a project** (see "Matins responsory pools" and its
eleven follow-up entries, 2026-09-04/05) — the per-Common-category pool (19/19 real categories) is
also done; see that same first entry. Extended the Missa check to the other 6 already-absent books
too (Josue, Judices, Ruth, 1-2 Chronicles) — genuinely zero citations anywhere, Horas or Missa.
Ezra came up with one thin, ambiguous lead not acted on: `missa/Latin/Tempora/Pasc7-2.txt`'s
Introit cites `!Esdr 2:36 2:37`, but the actual Vulgate text there (checked against
`data/scripture/1esdr-2.yml`) is a plain census list that doesn't match the Introit's own text at
all ("Accípite jucunditátem glóriæ vestræ... qui vos ad cœléstia regna vocávit"). User's own
explanation (2026-09-05), plausible and probably right, though unprovable without a text we don't
have: Vulgate numbering is genuinely ambiguous here — "1/2 Esdras" is canonical Ezra/Nehemiah, but
there's also a separate, non-canonical Greek "Esdras A" (the Vulgate appendix's own "3 Esdras,"
sometimes itself called "1 Esdras" in Septuagint/Orthodox numbering) — a different text entirely
that this project has never digitized. The reference engine's own citation never disambiguates
("Esdr" bare, no "3" anywhere else in the source to compare against), so this can't be confirmed
outright, but it cleanly explains why the citation doesn't match our canonical Ezra text. Left
both `ezra` and `neh` unseeded rather than force an uncertain match. One small, unrelated loose end
noted along the way
and not addressed here: the bible-plan TSV's own content still isn't covered by a table-driven
test the way other bulk-imported content stores are (see "Bible-plan TSV — bulk conversion done"
below) — `tests/propers/bible-plan.test.ts` tests specific real dates ad hoc, not the full imported
set.

### Matins responsory pools: `ruth` — the first composition with no liturgical precedent at all (2026-09-05)

User-directed: compose Ruth's own responsory from Ruth 1:16-17 ("thy people shall be my people,
and thy God my God") on the verse's own merit, even though — unlike `deut`/`rom`/`bar` — this text
has *no* liturgical precedent anywhere in the reference source at all (no Nuptial Mass file even
exists there to check against). Flagged that distinction directly before proceeding; user confirmed
it explicitly, aware of the difference from the three prior compositions (all of which recombined
text already used liturgically elsewhere, just never in Matins).

Since the project's usual bulk-imported `data/scripture/ruth-1.yml` carries no accents, the Latin
was hand-accented rather than transcribed — but cross-checked word-by-word against already-accented
occurrences of the same words elsewhere in the reference corpus (`quocúmque`, `pópulus tuus`,
`páriter`, `ábeam`, `relínquam`, `moráta`, `morábor`, `móriar`, `moriéntem`, `suscéperit`,
`sepultúræ`) rather than guessed freehand from scratch.

66 of 72 bible-plan books now seeded. `npm test` (2029 passed) and `tsc --noEmit` both pass;
live-verified through `getBiblePlanReadings` (`sunday-after-ascension`/Saturday resolves real,
correctly R./V.-formatted text in both languages for `Ruth 3-4`). Canonical-key audit re-run, clean
(`ruth` has no `BOOK_ALIASES` entry). Remaining 6: Josue, Judices, 1-2 Chronicles, Nehemiah (no
citation anywhere at all) plus `ezra` (one thin, ambiguous near-miss, deliberately left unseeded —
see the entry above).

### Matins responsory pools: `2esdr` (Nehemiah) — correcting a mis-diagnosis (2026-09-05)

The entry above's "Nehemiah: no citation anywhere at all" was wrong — Nehemiah had been bundled in
with `ezra`'s "one thin, ambiguous near-miss" reasoning without ever getting its own Missa check, the
same check that had already found and closed `bar` a few entries earlier. It has a real, unambiguous
citation: Ember Wednesday of September's own Lectio and Communio both quote Neh 8
(`missa/Latin/Tempora/093-3.txt`, `!Neh 8:1-10` / `!Neh 8:10`), genuine accented Latin, matching
English in the sibling `English/Tempora/093-3.txt`.

User-directed: R. from v.3 ("Et legit in eo apérte in platéa... et aures omnis pópuli erant eréctæ
ad librum" — Esdras reading the law plainly, all ears attentive), V. from v.9 ("Dies sanctificátus
est Dómino, Deo nostro: nolíte lugére et nolíte flere" — user's pick, offered against v.6's "Amen,
Amen" and v.10's "Ite, comédite pínguia," for the contrast between the plain reading and the
proclaimed holy day of no mourning). Both Latin lines taken verbatim from the same accented Missa
Lectio; English from the project's own verified `src/data/scripture/2esdr-8.yml`, not retranslated.
Keyed `2esdr:` in the pool (not `neh:`) since that's what `BOOK_ALIASES` maps `neh` to.

67 of 72 seeded. Remaining 5: Josue, Judices, 1-2 Chronicles (no citation anywhere at all) plus
`ezra` (one thin, ambiguous near-miss, deliberately left unseeded — see the entry above).

### Matins responsory pools: `1esdr` (Ezra) — composed, no liturgical precedent (2026-09-05)

User-directed: compose Ezra's own responsory from 1 Esdras 8:21-23, on the verse's own merit — like
`ruth`, not like `bar`/`2esdr`, since the one candidate lead for this book (`Pasc7-2.txt`'s Introit,
see the earlier `ezra`/`neh` entry) turned out to be a text mismatch, not a real match to build from.

v.22's antithesis ("the hand of our God is upon all them that seek him in goodness: and his power and
strength, and wrath upon all them that forsake him") is the strongest material, but the Vulgate's own
clause order puts the wrath half last — a sour note to end the repeated (`*`) clause on, especially
right before v.23's "it fell out prosperously." Edited: swapped the two halves of v.22 so the hopeful
clause is the one repeated after V. — same words, no invention, just reordered for the R./V./R. shape.
V. is v.23 unaltered.

Source (`1esdr-8.yml`) has no accents, like `ruth-1.yml`; Latin hand-accented, cross-checked
word-by-word against already-accented occurrences elsewhere in the corpus (`impérium`, `fortitúdo`,
`derelínquunt`, `bonitáte`, `jejunávimus`, `próspere`). Keyed `1esdr:` (`BOOK_ALIASES` maps `ezra` to
it). `npm test` (2029 passed) and `tsc --noEmit` both pass; live-verified through
`getBiblePlanReadings` (`easter-sunday`/Monday resolves real, correctly R./V.-formatted text in both
languages for `Ezra 3-4`).

68 of 72 seeded. This closes every book with any real sourceable material — the remaining 3 (Josue,
Judices, 1-2 Chronicles) have no citation anywhere, Horas or Missa, and no scripture text of their
own has been composed from on its own merit the way `ruth`/`ezra` were, since that hasn't been
directed for them yet (Josue itself got exactly this treatment shortly after — see the dated entry
below).

### Matins responsory pools: `jos` (Josue) — composed, no liturgical precedent (2026-09-05)

User-directed: compose Josue's own responsory from Josue 24:15-16, on the verse's own merit — like
`ruth`/`1esdr`, since Josue has no citation anywhere in the reference source at all, Horas or Missa
(re-confirmed live: every "Josue" hit in the corpus is an incidental mention of the person inside
unrelated books' readings — Ambrose's *De Officiis* homily, 1-2 Kings — never the Book of Josue's
own text).

The verse pair is already a two-voice exchange — Joshua's challenge to Israel ("choose you this day
... but as for me and my house, we will serve the Lord") and the people's reply — mapping directly
onto R./V. form with no invented content, just split at the natural seam. Asterisk placed before
"Ego autem", so the full choice-clause is R. and the famous declaration is the repeated close; V. is
v.16 kept whole, including "Respondítque pópulus, et ait" (not trimmed for the change of speaker,
since R./V. form doesn't make that redundant here — the narrative frame still matters).

Source (`jos-24.yml`) has no accents, like `ruth-1.yml`/`1esdr-8.yml`; Latin hand-accented,
cross-checked word-by-word against already-accented occurrences elsewhere in the corpus (`vidétur`,
`Dómino`, `elígite`, `hódie`, `potíssimum`, `servíre`, `debeátis`, `serviérunt`, `Mesopotámia`,
`habitátis`, `serviémus`, `Respondítque`, `pópulus`, `relinquámus`, `serviámus`, `aliénis`); `óptio`
left unaccented (2 syllables, no precedent either way, but this project's own accenting convention
never marks a 2-syllable word). `Amorreorum` kept in the spelling `jos-24.yml`'s own bulk import
uses (no `h`, single `r`) rather than the Clementine `Amorrhæórum` spelling found elsewhere in the
corpus (Ps 134/135), since that's this book's own source-text tradition, not a typo to "fix" toward
the more familiar spelling.

Keyed `jos:` (`BOOK_ALIASES` maps `josh` to this canonical id) — not `josue`, which would silently
miss every lookup. `npm test` (2029 passed) and `tsc --noEmit` both pass; live-verified through
`getBiblePlanReadings` (`easter-4`/Sunday resolves real, correctly R./V.-formatted text in both
languages for `Josh 1`).

69 of 72 seeded. Remaining 3 (Judices, 1-2 Chronicles) have no citation anywhere, Horas or Missa,
and no scripture text of their own has been composed on its own merit yet.

### Matins responsory pools: `jdc` (Judges) — composed, no liturgical precedent (2026-09-05)

User-directed: compose Judges' own responsory from Gideon's trumpet-and-pitchers stratagem at the
camp of Madian (Jdc 7:17-20), on the verse's own merit — like `jos`, since Judges has no citation
anywhere in the reference source at all, Horas or Missa.

Considered and set aside: the Song of Deborah's closing benediction (Jdc 5:31) has no natural second
verse to pair with it without inventing one; Samson's dying prayer (Jdc 16:28-30) was set aside as an
uncomfortable emphasis for a Matins responsory (a prayer for strength to kill himself with his
enemies). Gideon's cry won out as the single most recognizable line in the book, and — like
`ruth`/`jos` before it — the passage is already a command/fulfillment structure that maps directly
onto R./V. form: R. is v.17-18 (Gideon's command to the three hundred men); V. is v.19-20's
fulfillment, trimmed to its own core clause (dropped v.20's "in tribus locis... hydrias
confregissent" middle as restating what "conplodere lagoenas" already said), landing the versicle
directly on the iconic cry itself: "Gladius Domini, et Gedeonis."

Source (`jdc-7.yml`) has no accents, like `jos-24.yml`; Latin hand-accented, cross-checked
word-by-word against already-accented occurrences elsewhere in the corpus (`vidéritis`, `fácite`
— matching the "Hoc fácite" idiom's own established accenting — `ingrédiar`, `sectámini`,
`personáverit`, `castrórum`, `circúitum`, `clángite`, `conclamáte`, `tenuérunt`, `mánibus`,
`sonántes`, `clamaverúntque`, `Gládius`, `Gedeónis`, `Gedeóni`); `sinistris`/`dextris` left
unaccented, matching this corpus's own consistent practice for those two specific words even where
the general accenting rule would suggest otherwise. Also caught: the source's own English for v.18
is incomplete (drops "et conclamate, Domino et Gedeoni" entirely, ending mid-sentence) — supplied
the missing clause from the standard Douay-Rheims wording rather than reproducing that gap.

Keyed `jdc:` (`BOOK_ALIASES` maps `judg` to this canonical id). `npm test` (2029 passed) and
`tsc --noEmit` both pass; live-verified through `getBiblePlanReadings` (`easter-6`/Wednesday
resolves real, correctly R./V.-formatted text in both languages for `Judg 7`).

70 of 72 seeded. Remaining 2 (1-2 Paralipomenon/Chronicles) have no citation anywhere, Horas or
Missa, and no scripture text of their own has been composed on its own merit yet.

### Matins Nocturn 3 Gospel pericope — closed (2026-09-05)

Systemic gap found while reviewing the Ember-Friday/9-18 fix: a huge number of `nocturn-readings/
*.yml` files had an authored patristic homily (`isGospel: false`, a `source` field naming the
author/work) with no paired `isGospel: true` Gospel pericope — the homily discussing a Gospel text
that was never itself transcribed into the file. Nocturn 3 in this office is always a homily on
that day's own Mass Gospel; leaving the pericope unauthored meant the homily read with no context
for what it was actually commenting on.

Closed in two waves, since the first pass's detection method was itself too narrow (a regex over
the `source` field looking for an explicit citation) and missed real cases — either because the
citation was only in the file's header comment, phrased in a way the regex didn't match (a Latin
abbreviation, a parenthetical, a bare "Commentary on Luke" with no verse number), or not named
anywhere in the file at all:

- **Wave 1** (regex-detected, 64 files): every file where `source` itself named a Gospel
  book+chapter.
- **Wave 2** (structural re-scan — any `isGospel: false` + `source` entry in a file with zero
  `isGospel: true` entries, no regex): found 75 more candidates. Required real per-file
  verification against `../divinum-officium-reference` (checking the actual Lectio7-9/
  `[Evangelium]` structure), since several turned out to be Nocturn 2 content or Common-of-saints
  cross-references with no proper Gospel of the saint's own — ~46 authored, ~29 correctly left
  alone. A handful more (`st-aloysius-gonzaga.yml`, `st-francis-xavier.yml`,
  `st-john-baptist-de-la-salle.yml`, `st-joseph-calasanctius.yml`) were dropped from the wave-2
  task list by mistake and fixed directly during final verification.

Along the way, two more real bugs surfaced and were fixed in `src/hours/matins.ts`/
`src/calendar/day-label.ts`:
- A demoted temporal commemoration (e.g. a "Commemoratio Feriæ") was leaking a governing Sunday's
  own homily into an unrelated weekday's reading pool through a separate code path than the one a
  2026-09-03 fix had gated — affected 18 days/year.
- Reading pool order didn't match the day label's own commemoration order (temporal-then-
  sanctoral) — fixed to follow the same convention.
- A plain generic temporal commemoration (e.g. "Saturday in the 15th week after Trinity") was
  shown in the day label even when something else was already being celebrated, redundant with
  either an Ember day's own name or simply unnecessary noise alongside any winner — dropped
  unconditionally, not just for Ember days.
- A homily's own authored responsory (when distinct from its paired Gospel's) was never actually
  read by `gospelReadingPart` — silently dropped for every Gospel+homily pair with a responsory on
  the homily specifically (e.g. `post-pentecost-16.yml`'s Ambrose homily).

A handful of files were found to have **no Nocturn 3 content authored at all** (not just an
unpaired Gospel) — logged separately above under "Minimal-proof-done, bulk content pending", not
part of this closed item.

`npm test` (2029 passed) and `tsc --noEmit` both pass; live-verified on 2026-09-16/18/19 (the
Ember-September-into-saint cluster that originally surfaced this).

### Matins Nocturn 3 entirely unauthored — closed (2026-09-05)

6 files were missing Nocturn 3 entirely (not just an unpaired Gospel on an existing homily),
found while closing the Gospel-pericope gap above: `nativity-bvm.yml` (Matt 1:1-16 + St. Jerome,
Commentary on Matthew Bk. 1), `ss-cletus-and-marcellinus.yml`/`ss-soter-and-caius.yml`
(Common-of-Popes Nocturn 3, Matt 16:13-19 + Pope St. Leo the Great's Sermon 2 on the Anniversary
of his Election — reused verbatim from `st-martin-i.yml`, already authored this session with the
identical Common-of-Popes text), `ss-cornelius-and-cyprian.yml` (Common-of-Martyrs Nocturn 3,
Luke 21:9-19 + Pope St. Gregory's Homily 35), `ss-seven-brothers-rufina-and-secunda.yml` (a
proper override on top of its own Common, Matt 12:46-50 + Gregory's Homily 3, "Who is my
mother?"), and `st-anacletus.yml` (an alternate "rubrica 1570 aut rubrica 1617" variant of
`Sancti/07-13.txt`, carrying a real Nocturn 3 — Luke 14:26-33 + Gregory's Homily 37 — distinct
from the file's own modern-rubric default, which has none at all; of particular interest given
this app's own Monastic 1617 structural default). All sourced from `../divinum-officium-reference`
(the Common-of-Popes Gospel pericope again only found under `obsolete/missa/`, same pattern
as `st-isidore-of-seville.yml` above); `st-anacletus.yml`'s homily English has no reference-engine
translation for this specific Lectio7-9 split, so it was composed directly from the Latin and
marked draft.

`npm test` (2029 passed) and `tsc --noEmit` both pass; live-verified via `resolveOrdo` on all 6
saints' real dates.

### Common-of-a-Dedication Lauds — closed (2026-09-05)

The last open Lauds Common-category gap (9/10 -> 10/10). A 2026-08-28 note on this Common's
Vespers file claimed Lauds needed an OT canticle ("Canticum Ieremiæ") never transcribed anywhere
in this app — re-verified fresh against the live engine (Monastic Tridentinum 1617,
`command=prayLaudes`, `votive=C8`) across all 7 weekdays (11-09 through 11-15-2026) and found that
premise was wrong: no such canticle appears anywhere in the real scheme on any weekday. The
original note's numbers ("Ps 50/87/89 + Exodus canticle") match plain ferial Thursday's own Lauds
exactly (`Psalmi major.txt`'s `[Monastic Laudes] (feria 5)`) — the 2026-08-28 session evidently
captured that instead of actually passing `votive=C8`.

The real scheme: 3 proper-antiphon groups (Ps 50/5/35, this Common's own text) plus a canticle
slot whose *antiphon* is fixed to this Common ("Bene fundáta est...") but whose underlying
canticle *text* simply follows the plain ferial weekday rotation — Isaiah/Ezechias/Anna/
Moses-Exodus/Habacuc/Moses-Deuteronomy-split/Three-Young-Men — all 7 already authored
(`hours/lauds-canticles.ts`). Required a small mechanism change since this is the first Common
whose canticle id isn't fixed: `LaudsPsalmodyOverride.canticle.id` (`hours/lauds-psalmody-
overrides.ts`) is now optional, and `hours/lauds.ts`'s `resolvePsalmody` splices in the day's own
ferial canticle id/split when an override omits it, keeping the override's own antiphon. No other
category is affected (all author a fixed id already).

`npm test` and `tsc --noEmit` both pass.