docs: add a real backlog (TODO.md)

Was scattered across code comments and session summaries with no
single place to check what's actually left open -- consolidated here,
with concrete lists (which duplex-majus+ saints still need a Lauds
psalmody override, which already-authored Benedictus antiphons are
unverified against Monastic 1617 vs. the Roman rite they may have
actually been sourced from) rather than vague pointers.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

TODO.md

@@ -0,0 +1,119 @@
+# 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."
+
+## Content gaps — mechanism built, data incomplete
+
+### Lauds psalmody overrides (duplex-majus+ per-feast psalms/antiphons/canticle)
+
+Mechanism: `src/hours/lauds-psalmody-overrides.ts`, gated in
+`hours/lauds.ts`'s `getPsalmodyOverrideFor`. Only fires for a saint ranked
+duplex-majus or higher (per direct instruction) or a named temporal feast
+(`marian-saturday`, `christ-the-king`).
+
+**Authored (3):** `st-lawrence`, `marian-saturday`, `christ-the-king`.
+
+**Not authored yet (32 duplex-majus+ saints, sorted roughly by calendar
+order)** — each needs a live `command=prayLaudes` query on the saint's own
+date, Monastic Tridentinum 1617, same method as the three done so far:
+
+conversion-of-st-paul, chair-of-st-peter-at-rome, apparition-of-st-michael,
+st-matthias, st-benedict\*, st-mark, ss-philip-and-james, st-john-before-the-latin-gate,
+st-john-apostle (Dec 27, distinct from st-john-before-the-latin-gate May 6),
+st-barnabas, nativity-of-st-john-the-baptist, ss-peter-and-paul, chair-of-st-peter-at-antioch,
+st-james-the-greater, st-martin-of-tours, holy-innocents, st-thomas-apostle,
+all-saints, all-saints-of-the-benedictine-order\*, st-andrew, st-stephen-protomartyr,
+ss-simon-and-jude, st-luke, st-matthew, st-michael\*, st-maurus-abbot\*,
+ss-placid-and-companions\*, st-scholastica, st-bartholomew, st-peter-ad-vincula,
+assumption, nativity-bvm, immaculate-conception.
+
+\* = `propers: null` in the saint's own record — needs collect + antiphon
+authored too, not just the psalmody override, before any of this is useful.
+
+### Lauds office-bundle overrides (capitulum/responsory/hymn/versicle)
+
+Same override id, separate files (`lauds-capitulum-<id>.yml` etc. in
+`data/propers/common/`) — falls back to the plain weekday bundle when
+absent, *independently* of whether the psalmody override exists. Right now
+**only `christ-the-king` has this** — St. Lawrence's own day shows his
+proper psalms but the plain ferial capitulum/hymn, which is inconsistent
+with what the live engine actually shows for him (`2 Cor 9:6`, his own
+hymn). Worth fixing even before tackling the other 32.
+
+### Benedictus antiphons for temporal (non-sanctoral) days
+
+`<id>-benedictus-antiphon.yml` in `data/propers/temporal/` — only
+`marian-saturday` and `christ-the-king` exist. Every plain numbered Sunday
+(52 of them) and feria is `missing`. Real practice has a proper antiphon
+for every one of these; not started, and it's a big pull if done exhaustively.
+
+### Octave collects
+
+`getDayCollects` skips a `kind: 'octave'` commemoration entirely — no
+collect has ever been authored for an octave (the octave-reading pull only
+ever sourced Matins content). Every octave commemoration currently
+contributes zero text to Lauds'/Vespers' collect sequence.
+
+## Correctness risk — needs verification, not just authoring
+
+### Per-saint Benedictus antiphons may be sourced from the wrong rite
+
+`<saint>-antiphon.yml` files (17 of them, see below) were pulled during the
+original sanctoral-content sessions, months before Lauds existed to consume
+them. Building St. Lawrence's psalmody override surfaced, by accident, that
+his had been read from the **Roman/secular rite's `[Ant 1]`**, not Monastic
+1617's actual Benedictus antiphon for that day — two genuinely different
+texts. Fixed for Lawrence. **Not checked for any of the other 16**:
+
+apparition-of-st-michael, chair-of-st-peter-at-antioch, chair-of-st-peter-at-rome,
+conversion-of-st-paul, holy-innocents, nativity-bvm, nativity-of-st-john-the-baptist,
+ss-peter-and-paul, ss-philip-and-james, st-andrew, st-john-apostle,
+st-john-before-the-latin-gate, st-martin-of-tours, st-peter-ad-vincula,
+st-stephen-protomartyr, st-thomas-apostle.
+
+Each needs a live `command=prayLaudes` query on the saint's own day (Monastic
+1617) to confirm or correct — same check that caught Lawrence's.
+
+The other 12 duplex-majus+ saints with `propers` set have **no antiphon file
+at all yet** (st-scholastica, st-matthias, st-matthew, st-mark, st-luke,
+st-james-the-greater, st-bartholomew, st-barnabas, ss-simon-and-jude,
+immaculate-conception, assumption, all-saints) — these don't have the wrong-rite
+risk, they just don't exist yet.
+
+## Mechanism not built at all
+
+- **Vespers** (`src/hours/vespers.ts` is still a stub). First/Second Vespers
+  overlap (`DayWinner.vespersFrom`) already exists as a type but nothing
+  populates it for Vespers itself yet (only Compline uses `resolveEveningDay`
+  so far).
+- **Matins** (`src/hours/matins.ts` is still a stub) — needs its own
+  resolution path (reading-candidate pool, nocturn structure), not the
+  static parts-array pattern the other hours use.
+- **The "which Sunday governs this date" temporal-id algorithm** — the fixed
+  23rd-Sunday-before-Advent rule, Epiphany Sundays transferring to the end
+  of the liturgical year when they don't fit before Septuagesima, and the
+  Saturday-before-Septuagesima edge case (also commemorate the 6th Sunday
+  of Epiphany that day, in overflow years). Described in detail in an
+  earlier conversation, explicitly tabled, not started. Distinct from
+  Christ the King, which didn't need it (last-Sunday-of-October is direct
+  date math).
+- **Easter's own octave** — explicitly tabled ("easter needs a whole
+  different discussion").
+
+## Known, deliberate simplifications (not bugs — working as designed)
+
+- `getDayCollects`: each collect in a multi-collect day renders as its own
+  complete "Let us pray ... Amen." block, not the more compressed real form
+  (one "Let us pray," only the last collect's doxology).
+- Suffrage omission only checks `omitOnDouble` (+ Marian Saturday's own
+  Cross/BVM filter, + Christ the King's explicit id check). Real seasonal
+  suppression (e.g. during Advent/Lent) isn't modeled.
+- `data/calendar/temporal-categories.yml`'s season→category reconstruction
+  is from general knowledge of the pre-1955 tradition, not independently
+  verified against a primary source for every entry (flagged inline where
+  it matters).
+- St. Michael's Sep 29 date — kept as a deliberate simplification despite
+  the live engine showing nothing there across 6 tested years (documented
+  in `st-michael.yml` and `sanctoral-calendar.yml`).