docs(planning): W21 product-lead assessment

.planning/research/w21-product-lead-assessment.md

@@ -0,0 +1,349 @@
+# W21 — Product-Lead Assessment for v1.4.25
+
+**Author:** product-lead reviewer (read-only)
+**Branch:** `develop` (latest `51f23ef3`)
+**Scope:** strategic alignment between promised v1.4.25 release scope and the
+266 commits since `v1.4.24`. Sister reviewers (code, security, design,
+senior-dev, simplifier, i18n-runtime, dead-code) cover their respective
+domains; this report stays at product-strategy altitude.
+
+---
+
+## Executive summary
+
+v1.4.25 is the largest feature delta in the v1.4.x line, and every major
+promised theme is shipped on `develop`: onboarding rebuild (W14b), GLP-1
+Research Mode end-to-end (W19c with GROUND RULE 15 + MDR-acknowledgment +
+6-locale refusal probes), the W19d/e/f Wave-4b GLP-1 detail stack
+(side-effect taxonomy, cadence + compliance, EMA titration ladder),
+source-priority two-axis (W8c), Withings Activity + Sleep v2 (W17b/c),
+workout-batch ingest (W16b), PR detection worker + badge (W16c), W14c
+native FR/ES/IT/PL Coach prompts, dashboard polish (W20a), and the W15
+hygiene cluster (380 dead i18n keys, dead prompt constants, W7d
+hardening). PR #168 carries the work and `Lint+Test`, `Integration`,
+`Dependency Audit`, `Secret Scanning`, both Docker builds are all green;
+`e2e` was still IN_PROGRESS at the time of writing.
+
+**Headline verdict: v1.4.25 needs 1 reconcile action before tag.**
+
+The single Critical is the CHANGELOG: the W11 release-prep commit
+`cb07d5c` froze the changelog at the Wave-1-to-3 surface, and 60+
+subsequent Wave-4-and-5 commits (the W14b/c, W16a/b/c, W17a/b/c, W19a-f
+families, plus W15/W18/W20-rest and the Fix-G/H integration hot-fixes)
+have not been written up in CHANGELOG.md. This was already in the
+Session-2 handoff as the explicit `W22 Release-redo` step ("Update
+CHANGELOG.md to reflect Wave 4-5 additions (~500 lines total)"). The
+content is fully traceable from the phase reports in `.planning/`; the
+work is editorial, not engineering. Until that lands the public release
+artifact misrepresents what shipped.
+
+Everything else is strategically aligned. Marc-Voice is intact across
+both the CHANGELOG and the new user-facing copy. The MDR boundary is
+preserved with three concentric defences (chart gating, dialog
+acknowledgment, Coach refusal). The OpenAPI hard-flip is in place, so
+the v1.5 iOS sprint can rely on contract stability. The migration train
+(0043 → 0059) is forward-only and additive.
+
+---
+
+## Critical
+
+### C-1 — CHANGELOG.md is frozen at the Wave-3 surface and omits the entire Wave-4-and-5 delivery
+*Severity: Critical (blocks the tag — public release artifact misrepresents what shipped)*
+
+The current CHANGELOG.md `## [1.4.25]` section (lines 3-388) was written
+in `cb07d5c chore(release): v1.4.25` during the W11 release-prep agent.
+Since then 60+ commits have shipped across Wave-4 (W14b Foundation +
+Content, W19c Backend + Frontend + Safety, W19d, W19e, W19f) and Wave-5
+(W14a OpenAPI hard-flip, W14c native Coach prompts, W15 hygiene, W16a-c,
+W17a-c, W19a-b, W20-rest), plus Fix-G and Fix-H integration hot-fixes.
+None of this content is in the changelog.
+
+Cross-checking the spot-greps against CHANGELOG.md `## [1.4.25]`:
+
+| Promised feature | In CHANGELOG.md? |
+| --- | --- |
+| Onboarding rebuild (carousel, chips, source 4-card, baseline, done, welcome-back) | not mentioned |
+| GLP-1 drug-level chart + Research-Mode dialog | not mentioned |
+| GROUND RULE 15 (drug-level refusal, MDR boundary) | not mentioned |
+| 90-day acknowledgment staleness | not mentioned |
+| Side-effect taxonomy (21 entries × 5 categories) | not mentioned |
+| Cadence visualisation + compliance chips | not mentioned |
+| EMA titration ladder display | not mentioned |
+| Pen/vial inventory + 30-day clock (Migration 0056) | not mentioned |
+| W14c native FR/ES/IT/PL Coach prompts (replacement of REPLY LANGUAGE footer) | not mentioned (FR/ES/IT/PL still framed as "ride EN system prompts via REPLY LANGUAGE footer until native rewrites ship" — line 92-96) |
+| W17b Withings Activity sync | listed as deferred to v1.4.26 (line 355) |
+| W17c Withings Sleep v2 sync | listed as deferred to v1.4.26 (line 357) |
+| W17a Withings webhook secret as URL path-segment | not mentioned |
+| W16b POST /api/workouts/batch | not mentioned (line 354 still says "endpoint waits on iOS contract finalisation") |
+| W16c PR detection worker + badge | line 102-103 + line 351-352 still say "Detection worker deferred to v1.4.26" |
+| W14a OpenAPI drift-gate hard-flip | listed as deferred (line 342-344) |
+| W20a dashboard tile polish (single-line headings + inline trend arrow + baseline alignment) | not mentioned |
+| Medication inventory card + 30-day clock | not mentioned |
+| Migrations 0055-0059 | not mentioned (lines 17 + 332 cap at 0044-0054) |
+| PROMPT_VERSION already at 4.25.0 from W14c (pre-Wave-4) — GROUND RULE 15 is additive within 4.25.0 | not mentioned |
+
+The Session-2 handoff explicitly anticipated this: §"Remaining work →
+Wave 6 → **W22 Release-redo**: Update CHANGELOG.md to reflect Wave 4-5
+additions (~500 lines total). Version stays 1.4.25. New commit
+`chore(release): expand v1.4.25 with Wave 4-5 features` on develop.
+Push. Move PR #168 from Draft → Ready-for-review". W22 has not yet run.
+
+**Action**: dispatch W22 before Marc UAT. New `chore(release):` commit
+on `develop`. Test count update from the Session-2 progression table
+(2244 → ~3460-ish unit + ~165 integration at Fix-H) lands in the same
+edit. Re-categorise the Wave-4-5 items out of "Deferred to v1.4.26" and
+into Added / Changed / Fixed / Security / Refactor.
+
+---
+
+## High
+
+### H-1 — Test count footer in CHANGELOG is stale
+*Severity: High (Marc-Voice precision; the marathon outcome must read truthfully)*
+
+`### Tests` (lines 320-336) reads "2244 → 2652 passing unit tests
+(+408). Integration suite 140 → 174 (+34)." The Session-2 handoff
+records the actual progression at the Fix-H point as `~3400 unit + ~170
+integration` — over 700 additional unit tests landed in Wave-4-5
+(W19c-Safety alone added 1800+ refusal-probe assertions across 15 rules
+× 6 locales × 20+ probes; W14b added 45; W19d/e/f added 39+39+40; W16c
+added 40+4; W17b/c added 357 across the touched surface). Roll into the
+W22 CHANGELOG update.
+
+### H-2 — `Deferred to v1.4.26` section in CHANGELOG lists items that already shipped in Wave 4-5
+*Severity: High (release artifact is internally inconsistent; cross-references the C-1 reconcile)*
+
+Lines 337-368 still carry:
+- "OpenAPI drift-gate hard-fail flip (still warn-only after this release)" — W14a shipped the flip (`74bf608`, `41147bd`).
+- "Onboarding rebuild promoted from v1.5 (last release where new users see the v1.4.20-era onboarding...)" — W14b shipped it.
+- "Native FR/ES/IT/PL Coach + insights system prompts" — W14c shipped them.
+- "Personal-record detection worker (schema shipped this release; sweep-on-insert + nightly cron pending)" — W16c shipped both.
+- "Workout ingest endpoint" — W16b shipped it.
+- "Withings Activity sync routine" — W17b shipped it.
+- "Withings Sleep v2 routine" — W17c shipped it.
+- "414 dead i18n keys cleanup" — W15 shipped 380 (25 of the 414 verified live).
+
+If Marc tags from this state, the public release notes promise these are
+"deferred" while the binary actually contains them. W22 is the
+reconcile.
+
+---
+
+## Medium
+
+### M-1 — Drug-level chart-side staleness wiring deferred to v1.4.26 by the W19c-Safety phase
+*Severity: Medium (small surface, observed and documented; acceptable for v1.4.25 if Marc consents)*
+
+`src/lib/medications/research-mode-staleness.ts` is shipped with 18
+unit tests, but the conditional in `DrugLevelChart.tsx` to read
+`isAcknowledgmentStale()` is not wired — the chart today only gates on
+version mismatch, not on the 90-day clock. The W19c-Safety report
+documents this explicitly as a v1.4.26 cleanup. The MDR boundary is
+unaffected (the dialog still re-prompts on version bump, the Coach
+still refuses at the GROUND RULE 15 layer); the 90-day gate is a
+defence-in-depth ratchet. Acceptable to ship as-is but should be
+documented as known-deferred in the CHANGELOG Security or Deferred
+section, not silently omitted.
+
+### M-2 — `User.onboardingGoals` schema column + API persistence deferred to v1.4.26
+*Severity: Medium (W14b-Content stored goals only in localStorage; server cannot personalise dashboard from onboarding selections)*
+
+W14b-Content's goals chip-picker holds the selection in localStorage
+keyed by `userId`. The v1.4.20 dashboard-widgets seed feature (which
+the brief floated as the consumer) cannot read it from the server.
+Marc-acceptable per the W14b-Content deviation §1 (path explicitly
+listed in the brief as second-preferred); call out as a v1.4.26
+follow-up in CHANGELOG so future readers don't expect server-side
+goal persistence yet.
+
+### M-3 — W11a multi-arch image lowercase fix only landed in Fix-H
+*Severity: Medium (resolved, but only after PR #168 had been opened — third release in a row that needed a CI fix-up)*
+
+Fix-H commit `a6f94cf` resolved the GHCR push regression by lowercasing
+`IMAGE_NAME`. Was caught before Marc tagged but underscores Marc's
+existing concern from his memory ("Coolify auto-deploy fix" in v1.4.21,
+v1.4.22, v1.4.23 outcome memos). W20-rest commit `e23dd28` shipped the
+explicit `vars.COOLIFY_AUTO_DEPLOY=on/off` maintainer toggle so the
+silent-miss pattern can't recur. Strategic resolution is in place;
+M-severity only because the regression was new and the v1.4.25 release
+narrative ("multi-arch image lands clean") could read as smoother than
+it actually was.
+
+### M-4 — Session-1 handoff's W12/W13 deliverables (post-tag) carry items that aren't acknowledged in Session-2's W24 plan
+*Severity: Medium (operational continuity)*
+
+Session-1 §2 W12-W13 lists items (FUNDING.yml decision, social-preview-
+image upload, branch-protection required-status-checks flip, sister-repo
+healthlog-docs + healthlog-landing version-pin updates, `docs/audit/
+v1425-summary.md`, 0-10 score per area, Codex audit prompt, `v15-ios-
+handoff.md`). Session-2's W24 plan mentions these but does not break
+them out into a checklist the post-tag agent (or Marc himself) can
+follow. Not a blocker — Marc can drive these by hand — but worth
+calling out so the post-tag agent has a single source of truth.
+
+### M-5 — i18n FR/ES/IT/PL Coach prompts are now LLM-quality drafts with structural-coverage tests, not native
+*Severity: Medium (strategic-alignment with Marc's i18n promise; not a regression — the W14c agent landed the matrix + 1800+ refusal probes per Marc's reversal of the original "EN body + REPLY LANGUAGE footer" plan)*
+
+W14c shipped the native body per locale (replacing the REPLY LANGUAGE
+footer for FR/ES/IT/PL) but did so with LLM-quality drafts because Marc
+cannot review safety-critical FR/ES/IT/PL prose himself. The
+MaintainershipBanner acknowledges this. Strategically aligned with
+Marc-confirmed decision 1 in Session-2 §"Marc-confirmed strategic
+decisions". W22 CHANGELOG must surface this honestly — currently the
+shipped state is "AI-initial bundles riding the EN coach with
+MaintainershipBanner", but the actual state is "native LLM-drafted
+bodies per locale with the matrix + 1800+ refusal-probe assertions for
+safety coverage". Different positioning for a release public artifact
+vs the AI-as-author criticism. Marc's directive is "never expose
+Claude / AI-as-author" — the CHANGELOG should describe what shipped
+(native per-locale Coach prompts with maintainership banner) without
+explicitly billing them as "AI-drafted".
+
+---
+
+## Low
+
+### L-1 — CHANGELOG line 14 "AI-translated locales (FR / ES / IT / PL) ship behind a maintainership banner"
+*Severity: Low (Marc-Voice nit — the framing "AI-translated" is honest but the phrase pattern matches Marc-memory criticisms of AI-as-author exposure)*
+
+Recommend rewording to "community-maintainable" or "drafted-with-LLM-
+seeding-and-structural-matrix-coverage" or similar Marc-Voice
+formulation. The existing `<MaintainershipBanner>` UI copy speaks for
+itself; the changelog doesn't need to call out the translation
+provenance. (Marc-memory: "never expose Claude / AI / agent / phase /
+wave / marathon in user-facing artefacts" — CHANGELOG is user-facing.)
+
+### L-2 — W19c-Frontend dialog cites EMA EPAR + Schneck/Urva 2024 DOI
+*Severity: Low (no PII risk; literature citations are public; flagged for awareness only)*
+
+Cites are clinical references, not PII. Acceptable per Marc-memory PII
+gate which targets Marc's personal data (name, health figures, BD-
+Zielbereich values, measurement counts) — not pharmacovigilance
+literature. No action.
+
+### L-3 — `personal_records` `NULLS DISTINCT` index requires application-level pre-flight `findFirst`
+*Severity: Low (W16c shipped the application guard; Migration 0055-or-later can drop the guard)*
+
+W16c phase report §"Key finding" explicitly documents the
+application-level idempotency guard. Tracked in the v1.4.26 backlog.
+No action for v1.4.25.
+
+### L-4 — Two-axis source-priority Audit-log writes are mentioned in CHANGELOG (line 280-281) but the Audit-log surface itself is a v1.4.26 dead-code candidate
+*Severity: Low (cross-reference v1.4.26 backlog P4-6 + P6-10 — Marc decision pending on `/api/audit-log`)*
+
+The `AuditLog` row is written; the `/api/audit-log` GET is not consumed
+by any UI. Marc-decision needed per W20-rest P6-10 recommendations
+table. Track as v1.4.26.
+
+### L-5 — `MOOD_LABEL_KEYS` already-shipped tag on P6-11 (W20-rest report) is accurate but not reflected in CHANGELOG
+*Severity: Low (already in the changelog under "mood verbal labels" refactor bullet — closed)*
+
+CHANGELOG line 304 mentions it under refactor; the W20-rest closure note
+is consistent. No action.
+
+### L-6 — `MeasurementSideEffect` Prisma block triggered a one-time `prisma format` whitespace re-flow
+*Severity: Low (cosmetic; W19d phase report deviation §2 explicitly documents)*
+
+No functional impact. No action.
+
+---
+
+## Scope-delivery matrix
+
+| Wave | Feature | Promised | Shipped on `develop` | Verdict |
+| --- | --- | --- | --- | --- |
+| W14a | OpenAPI drift-gate hard-flip | Yes (Marc-directive — pulled forward from v1.4.26) | Yes (`74bf608`, `41147bd`) | Shipped |
+| W14b-Foundation | User.onboardingStep migration + OnboardingShell + nested routes + step API + i18n key surface | Yes | Yes (5 commits ending at `489d0da`) | Shipped |
+| W14b-Content | Carousel + goals chips + source 4-card + baseline + done + welcome-back + entry-point swap | Yes | Yes (5 commits ending at `b896622`) | Shipped (server-side goal persistence deferred to v1.4.26 — see M-2) |
+| W14c | Native FR/ES/IT/PL Coach prompts + safety matrix + 1800+ refusal probes | Yes | Yes (5 commits ending at `75fce6c`) | Shipped (M-5 framing note) |
+| W15 | 414 dead i18n keys + dead prompt constants + W7d hardening + Cat-C typo | Yes | Yes (5 commits ending at `fcda115`) — 380 removed, 25 verified-live and kept | Shipped |
+| W16a | iOS-17/18 HK long-tail enum extensions + VO2 chart-row helper + workout dedup TODO | Yes | Yes (3 commits ending at `bff13e7`) | Shipped |
+| W16b | POST /api/workouts/batch typed ingest | Yes | Yes (`62e4b1d` → `5a7d252`) | Shipped |
+| W16c | PR Detection worker + pg-boss queue + batch route hooks + badge + opt-in toggle | Yes | Yes (6 commits ending at `223b8a9`) | Shipped |
+| W17a | Withings webhook secret as URL path-segment | Yes | Yes (`a1ffa49`) | Shipped |
+| W17b | Withings Activity sync routine | Yes (Marc-directive — pulled forward) | Yes (Migration 0055, 6 commits at `dab7de3`) | Shipped |
+| W17c | Withings Sleep v2 routine | Yes | Yes (same 6-commit train as W17b) | Shipped |
+| W18 | 12 Low + apply-with-care items from W10 review | Yes | 8 / 12 shipped (4 deferred and tracked in v1.4.26 backlog) | Shipped (acceptable) |
+| W19a | EMA drug knowledge layer + drift-guard | Yes | Yes (`cee5bf5`, `da73e06`, `45bbfe4`) | Shipped |
+| W19b | Pen/vial inventory + 30-day clock + medication inventory card + cron | Yes | Yes (Migration 0056, `570b14d` → `7f133a1`) | Shipped |
+| W19c-Backend | Migration 0058 + glp1-pk module + research-mode endpoint | Yes | Yes (4 commits ending at `cf27df4`) | Shipped |
+| W19c-Frontend | MDR-acknowledgment dialog + drug-level chart + Settings toggle | Yes | Yes (4 commits ending at `32c1a41`) | Shipped |
+| W19c-Safety | Coach GROUND RULE 15 + 6 YAMLs + 1800+ refusal probes + 90-day staleness helper | Yes | Yes (4 commits ending at `bfd129f`) — chart-side staleness wiring deferred (M-1) | Shipped (M-1 wiring deferred) |
+| W19d | Side-effect taxonomy (21 entries × 5 categories) + API + UI section | Yes | Yes (Migration 0059, 4 commits ending at `7dfc212`) | Shipped |
+| W19e | Cadence visualisation + compliance chips | Yes | Yes (4 commits ending at `71cc05a`) — reused existing pg-boss + dispatcher per Marc-correction | Shipped |
+| W19f | EMA titration ladder display (read-only reference) | Yes | Yes (4 commits ending at `86dbff0`) — observational copy only, MDR-safe | Shipped |
+| W20a | Dashboard top-tile polish (single-line headings + inline trend arrow + baseline alignment) | Yes (Marc-asked mid-session) | Yes (3 commits ending at `a7cc5de`) | Shipped |
+| W20-rest | 11 P6 polish items | Yes | 3 / 11 shipped (Pearson math, Coolify toggle, translation-issue template); 8 deferred to v1.4.26 with maintainer recommendations | Shipped (acceptable) |
+| W8c | Source-priority two-axis (metric × deviceType) | Yes (Session 1 W8 — already in current CHANGELOG) | Yes (Migration 0051, `f05c55f` → `53278cc`) | Shipped (already in changelog) |
+| Fix-G/Fix-H | CI / integration / GHCR hot-fixes pre-tag | Internal | All green except the e2e dashboard insight-text + chart-tick locator, addressed via Fix-I `03bf21e5` + `d22fc212` | Shipped |
+| W22 (release-redo) | CHANGELOG expand to reflect Wave-4-5 + bump test counts + reshuffle "Deferred to v1.4.26" | Yes | **Not yet run** — see C-1 | **Reconcile required** |
+| W23 (STOP for Marc UAT) | Marc-only | n/a | n/a | Awaits Marc |
+| W24 (post-tag deploy + deliverables) | Demo redeploy, sister-repo updates, `docs/audit/v1425-summary.md`, 0-10 score, Codex audit prompt, `v15-ios-handoff.md` | Yes | **Not yet run — post-tag** | Awaits tag |
+
+---
+
+## v1.5 iOS readiness statement
+
+The iOS sprint has the contracts it needs to start Day 1.
+
+**Locked server contracts:**
+
+- `POST /api/measurements/batch` — Apple Health bulk ingest (v1.4.23) with `inserted | duplicate | skipped` per-entry envelope, idempotency wrapper, rate-limited.
+- `DELETE /api/measurements/by-external-ids` — HealthKit deletion-sync (v1.4.25 W4b).
+- `POST /api/workouts/batch` — HKWorkout-aligned typed ingest with 20-sport-type union + GeoJSON LineString routes (v1.4.25 W16b).
+- `POST /api/devices` — APNs token registration with paired `apnsEnvironment` (v1.4.23).
+- `POST /api/onboarding/step` — wizard progression with rate limit + audit (v1.4.25 W14b-Foundation).
+- `User.sourcePriorityJson` two-axis resolver with `pickCanonicalSource()` (v1.4.25 W8c) — iOS sets Apple Health as the cumulative + sleep + HRV + RHR favourite; Withings stays the point-measurement default.
+- `Measurement.deviceType` enum (v1.4.25 W8d) with iOS-17/18 HK long-tail mappings ready to extend.
+- `MedicationInventoryItem` (v1.4.25 W19b) — pen/vial 30-day clock available for iOS pen-scanner integration.
+- `Workout` + `WorkoutRoute` (v1.4.25 W8d Migration 0053) — table waits for iOS workout sync.
+- `PersonalRecord` (v1.4.25 W8d Migration 0054) — detection worker (W16c) writes the table now; iOS can read via `GET /api/personal-records?metricType=X`.
+
+**Safety guarantees iOS inherits at Tag-1:**
+
+- GROUND RULE 9 (Coach refuses GLP-1 dose recommendations) + GROUND RULE 15 (Coach refuses drug-level estimates with EU MDR + MDCG 2021-24 cites). Refusal-probe matrix in CI on every push — 1800+ assertions × 15 rules × 6 locales × 20+ paraphrasings — so iOS prompt regressions catch in CI not on-device.
+- OpenAPI drift-gate is now hard-fail (`41147bd`). iOS Swift codegen has confidence that develop's `docs/api/openapi.yaml` is in lockstep with the Zod registry. Any route signature change between this tag and the first iOS DTO regen surfaces in CI before iOS reaches it.
+- `requireAuth()` narrowed-scope path (v1.4.25 fix line 250-256) — iOS Bearer tokens with route-specific scopes (e.g. `medication:ingest`) now route through cleanly when the declared scope is in the token set, while wildcard tokens (`["*"]`) keep the existing v1.4.24 fail-closed posture for unscoped routes.
+
+**Apple Health migrations 0051-0055 (the iOS-readiness window):**
+
+- 0051 `measurement_device_type` — added the column.
+- 0052 `apple_health_enum_extensions` — added `ENVIRONMENT_AUDIO_EXPOSURE`, `HEADPHONE_AUDIO_EXPOSURE`, `TIME_IN_DAYLIGHT`.
+- 0053 `workout_and_route` — Workout + WorkoutRoute tables.
+- 0054 `personal_record` — PersonalRecord table + dedup index (NULLS-DISTINCT documented; W16c applies application-level guard).
+- 0055 `measurement_sleepstage_composite` — widened the Measurement unique index to include `sleep_stage` with `NULLS NOT DISTINCT` (Postgres 16 native).
+
+All five are present, additive, forward-only, and covered by integration tests (`tests/integration/withings-sleep-stage-composite.test.ts`, `tests/integration/pr-detection-end-to-end.test.ts`, etc.). Migration 0060 stays free for the next wave.
+
+**Net iOS readiness:** v1.4.25's contract surface is tighter and more
+complete than the v1.4.23 iOS-prep promise. The iOS Swift sprint can
+proceed against `develop` HEAD (after Marc tags). The W19 series adds a
+GLP-1 medication-tracking surface that iOS can wrap natively without
+the server moving underneath it.
+
+---
+
+## Closing
+
+Strategic alignment is intact. The shipped scope matches Marc's
+Session-1 + Session-2 directives and Marc-confirmed strategic decisions.
+The MDR boundary, the regulatory cites, the refusal probes and the
+"reference not advice" framing on titration are all conservative,
+defensive, and consistent with HealthLog's clinical-but-warm voice. AI
+Insights remain HealthLog's differentiator and the W19 series deepens
+it without diluting it — Coach gains a refusal layer rather than a
+prescription layer, and the four locale Coach bodies pick up structural
+safety coverage (1800+ probes) rather than thin REPLY LANGUAGE footers.
+
+**One reconcile action before tag:** W22 — expand CHANGELOG.md to
+reflect the Wave-4-and-5 delivery. The content is already in the phase
+reports under `.planning/`. Once W22 runs, Marc can UAT and tag without
+the public release artifact misrepresenting the contents of the
+binary.
+
+After tag, Session-2's W24 list (Demo redeploy, sister-repo version pins,
+`docs/audit/v1425-summary.md`, 0-10 score per area, Codex audit prompt,
+`v15-ios-handoff.md`) feeds the post-tag agent or Marc directly.
+
+No critical-pre-v1.5 blockers detected beyond W22.