LevSMS — Full Code + Sheet Audit

Verdict: the system is healthy and pilot-ready. No bugs found. The stale-davening bug that started this work is gone. The audit surfaces only cleanup + 3 decisions, none blocking: 2 dead functions (removed this pass), 4 vestigial sheet rows (auto-pruned by the cleanup already shipped), and 2 orphaned tabs + 1 dark delivery log that are leftover from the original architecture — harmless, but yours to keep, wire, or archive.

1. Code audit — clean

103 live functions, all 11 top-level constants used. No stale constants, no broken references.
Removed this pass (0 call sites): relTime_ (orphaned when ZMANIM switched from relative "in 23m" tags to absolute "as of" times) and parseDate_ (never called). Committed 7293614, live @32.
Kept: runWeeklyReview — looks unused but it's a UI-runnable wrapper for levSmsWeeklyReview_ (run it from the Apps Script editor). Same for seedShulTimesNW / installLevSmsTriggers.
Transparency holds: every user-facing reply renders from the sheet (SYSTEM_RESPONSES + APPROVED_RESPONSES). One framework key (BAD_AREA) lives only in the in-code default right now; it self-writes to the sheet on the next REFRESH.

2. Original → current architecture (the "old vs new")

The M1 rebuild changed where the davening schedule lives. This is the root of every "orphan" below — they're not bugs, they're leftovers from the first design.

Concern	Original design	Current (M1 rebuild)
Davening schedule source	Read live from the `SHUL_TIMES` tab (76 rows, rich schema)	Rendered from `NW_SCHEDULE` / `NW_SHULS` code constants + live Hebcal
Command routing	`COMMANDS` tab (Enabled / Public flags)	`resolveIntent_` + `CMD_SHORTCUTS` in code
Plag Mincha (KBH Bais Dovid)	Hand-entered weekly	Computed live: Hebcal plag − 10 (≈1 min earlier than shul board = safe, you never miss it)

3. Sheet audit — 10 tabs

Tab	Rows	Status	Note
INBOUND_LOG	311	LIVE	Every inbound text. Growing.
OUTBOUND_LOG	336	LIVE	Every reply + source + SENT status. Growing.
DELIVERY_LOG	311	DARK since 5/17	Twilio delivery callbacks. Sends don't request a `statusCallback`, so nothing's logged here since 5/17. Low impact — OUTBOUND_LOG already records SENT.
CONTACTS	41	LIVE	Phone / status / ZIP / tier. Real testers + you + the LevSMS line.
ZIP_REQUESTS	2	LIVE	Out-of-area ZIP requests queued for review.
SYSTEM_RESPONSES	13	LIVE	Framework messages (welcome, menu, help…). Missing `BAD_AREA` → self-heals on next REFRESH.
APPROVED_RESPONSES	17→13	LIVE	13 served commands, all updated 5/26. 4 vestigial rows (`MENU`, `YERUSHALMI`, `TANAKH`, `CHOFETZ` — frozen 5/14, pre-dating your "Daf+Mishna only" trim) are auto-pruned by the cleanup already shipped on next REFRESH.
COMMANDS	10	ORPHAN	Enabled/Public flags — not read by the router (routing is in code). Also incomplete (10 of ~18 commands). Decide: wire as a kill-switch, or archive.
SHUL_TIMES	76	ORPHAN	9 shuls, rich schema (season/conditional/shiur). Seeded but never read — superseded by code constants. Decide: promote to source-of-truth (more transparent, sheet-editable), or archive.
SETTINGS	4	LIVE	Key/Value config.

4. What's still needed (not built — by design)

M5 — drop-a-PDF/text auto-parse (the weekly schedule ingestion): gated on you adding ANTHROPIC_API_KEY to the LevSMS Script Properties.
M6 — admin reasoning + comic: storyboard + marketing prompts drafted; build not started.
WEBHOOK_SECRET anti-spoof: code is in place but dormant until you set the secret (optional hardening before a wider public launch).

5. Three decisions for you (none urgent)

SHUL_TIMES (76 rows): archive it, or make it the real schedule source so the davening times become sheet-editable (true "nothing hidden in code")? Today the rendered times are visible/editable in APPROVED_RESPONSES, so this is about source, not visibility.
COMMANDS (10 rows): wire the Enabled flag as a per-command kill-switch, or drop the tab? Right now the flags do nothing.
DELIVERY_LOG: add a statusCallback to sends so delivery/failures get tracked, or leave it (OUTBOUND already logs SENT)?

My recommendation: leave all three as-is for the pilot (they don't affect anyone texting in). Revisit SHUL_TIMES + DELIVERY only if/when you go past the pilot. Don't overcomplicate — confirmed by you this session.