Skip to content

TASKS.md

Latest commit

 

History

History
255 lines (192 loc) · 16.2 KB

TASKS.md

File metadata and controls

255 lines (192 loc) · 16.2 KB
Raw

TASKS.md

Agents: read this to find your task. Update status when you finish. Humans: add tasks here before starting an agent session.

One task = one logical change. If a task takes more than one session, split it.

See PLAN.md for the active sprint scope. Executable card details (anchors, RED/GREEN, verify) live in docs/plans/consolidated-audit-10x/ — do not invent scope beyond the card. See BUG_REGISTER.md for known defects outside this program.

In Progress

⚠ 2026-08-13 — this file's card program is CLOSED. See the PLAN.md banner before using anything below.

The active program is the RN store push (get the React Native successor into review on both stores), tracked in ~/.claude/plans/store-submission-runbook.md. It is not card-shaped and does not live here. The Phase-0/1/2/3 cards below are historical.

This repo's open items for that program (verified 2026-08-13):

Item Owner State
Run pnpm seed:reviewer-demo against production so the store reviewer sees real data (not empty lists) Owner ✅ DONE 2026-08-11 — the demo-5roles fire seeded 5 isolated demo clinics (reviewer-demo-clinic-*) + 5 prod Clerk users, verified against prod (vt_users/vt_shifts counts). Credentials CSV in the job dir; teardown owed after the G3 on-device verdict
Play App Signing SHA-256 → server/lib/well-known-assetlinks.ts + redeploy Agent blocked until the first AAB is uploaded
The doctor-gate + reviewer-seed contract suites were named by no workflow — incl. the doctor-gate test that pins the provenance contract the RN app depends on (the RN client sends source: "doctor_gate" in the request body; the server persists it as checkInSource, which is the field the test asserts) Agent ✅ CLOSED — coverage by 7a3ad3a (named files in integration-ops); the silent-skip hole in it closed separately: both suites are describe.skipIf(!dbReachable), so an unset or unreachable DATABASE_URL made them skip with the step still green. scripts/ci/db-integration-preflight.mjs now runs before them and refuses that
The rest of that class is still open. vite.config.ts excludes 10 entries from pnpm test (tests/migrations/** is a glob over 6 files); verified 2026-08-21 that none is named by any workflow: tests/migrations/**, restock.service, shift-chat-window, charge-alert-worker, code-blue-mode-equipment, equipment-scan-e2e, expiry-api, expiry-check-worker, returns-api. Two of those touch Code Blue / equipment scan. Corrected 2026-08-22: those six — the whole live-server half, including both safety-critical ones — now run in CI's live-server job (pnpm test:live-server), which the merge gate depends on, and which gates on the reported assertion count rather than the exit code alone. Three remain open: tests/migrations/**, restock.service and shift-chat-window are DB-only and need a different setup; they are NOT covered. tenant-pooling-isolation is the one deliberate exclusion (real DDL — run via test:rls-pooling). Do not read the row above as "the class is handled" Agent open
tenant:lint is --warn-only + --touched + continue-on-error, emitting ~208 findings of which ~197 are false positives (the heuristic never inspects the .where() chain) — a real clinicId leak would be indistinguishable from noise Agent open

⚠ Safety note — writing seed data to a production database. Verified against scripts/seed-reviewer-demo.ts on 2026-08-13, not assumed:

Property State
Tenant isolation ✅ writes only to its own namespaced clinic (reviewer-demo-clinic by default, or REVIEWER_DEMO_CLINIC_ID). The script carries an explicit guard against a typo'd/stale clinic id silently writing into a real clinic (scripts/seed-reviewer-demo.ts:60-62)
Idempotency ✅ every insert uses onConflictDoNothing / onConflictDoUpdate — safe to re-run
Synthetic data only ✅ demo persona + demo equipment; no real patient, staff, or clinic data is created or modified
Cleanup ⚠ partial. Rollback order is documented in docs/runbooks/code-blue-qa-walkthrough.md (children-first, all FKs are ON DELETE RESTRICT). But vt_audit_logs is append-only — once anything writes an audit row for the demo clinic, the clinic row itself becomes undeletable and leave-in-place is the only option

Before running: confirm REVIEWER_DEMO_CLINIC_ID is the demo clinic and not a real one, and take the routine production DB backup first. Accept that the demo clinic is effectively permanent once audited — it is unreachable by real users, so the cost is a stray row, not a risk.

Phase-0A cards below are COMPLETE (2026-07-12) — see banner and Completed. Do not re-execute.

Phase 0A cards (COMPLETED 2026-07-12 — reference only, do NOT execute)

⚠ 2026-07-28: T-05 · T-01 · T-02 · T-03 · T-04 all COMPLETED 2026-07-12 (commits b79f0819a, bb148cb3→b3c1f2e66, 364d21cfd→78c94841c, 4a1a75cc3→e451f0743, 332c311d2→9edf4845d; batch gate 492 files / 4552 tests, 0 fail — PROOF-logged). Cards retained for card-contract reference only.

All five cards COMPLETED 2026-07-12 (commits in the banner below). Full cards remain in docs/plans/consolidated-audit-10x/phase-0-1.plan.md for reference. After GREEN: pnpm test -- <RED file> && pnpm typecheck → log docs/audit/PROOF_ALIGNMENT_LOG.md → commit per card.

T-05: Pass QueryClient into initSyncEngine() (R-SY-01)

Priority: high · Tier: S +R · Linked plan: Phase 0A · foundational (do first)

What to do: Sole caller at src/hooks/use-sync.tsx invokes initSyncEngine() with no arg → queryClientRef stays undefined → post-replay invalidations, reconcile, and 401 cache-clear are no-ops. Pass useQueryClient() into initSyncEngine(queryClient).

Acceptance criteria:

  • RED then GREEN: tests/sync-engine-queryclient-wiring.test.ts (QueryClient passed; replay invalidates equipment; 401 clears cache)
  • Wiring only — no queue/circuit-breaker/emergency-cache changes
  • pnpm test -- tests/sync-engine-queryclient-wiring.test.ts && pnpm typecheck
  • code-reviewer pass before commit; proof logged

Files in scope: src/hooks/use-sync.tsx, src/lib/sync-engine.ts (read), test file
Files NOT in scope: sync queue semantics, SW cache paths, emergency endpoints

T-01: Code Blue outcome Cancel dismisses without ending session (R-CB-01)

Priority: high · Tier: S +R · Linked plan: Phase 0A

What to do: Cancel calls onClose("") → handleEndSession returns at the empty-outcome guard before closing the modal. Give Cancel a dedicated close that sets showOutcomeModal=false and does not call the end path.

Acceptance criteria:

  • RED then GREEN: tests/code-blue-outcome-cancel.test.tsx (modal gone; end mutation not called; focus returns)
  • No SSE/keepalive changes; no optimistic session end
  • pnpm test -- tests/code-blue-outcome-cancel.test.tsx && pnpm typecheck
  • code-reviewer pass before commit; proof logged

Files in scope: src/pages/code-blue.tsx, test file

T-02: Dock-Return + RFID sheets mount at page level (R-EQ-01/02)

Priority: high · Tier: S · Linked plan: Phase 0A

What to do: <DockReturnFlow> and <DockReturnNfc> sit inside inactive Readiness TabsContent → silent no-op on default tab. Move both to page level (mirror equipment-list).

Acceptance criteria:

  • RED then GREEN: tests/equipment-detail-dock-return-mount.test.tsx
  • Presentation/mount only — no custody-mutation change
  • pnpm test -- tests/equipment-detail-dock-return-mount.test.tsx && pnpm typecheck
  • Proof logged

Files in scope: src/pages/equipment-detail.tsx, test file

T-03: QR auto-decode last-scanned-wins exactly once (R-SC-01)

Priority: high · Tier: S · Linked plan: Phase 0A

What to do: Overlapping resolves can last-resolved-win the wrong tag and double-count scans. Monotonic token + stop scanner before await; apply resolve only if token still latest; increment once per applied scan.

Acceptance criteria:

  • RED then GREEN: tests/qr-scanner-race.test.tsx
  • Do not touch classifyEmergencyEndpoint / offline emergency block
  • pnpm test -- tests/qr-scanner-race.test.tsx && pnpm typecheck
  • Proof logged

Files in scope: src/components/qr-scanner.tsx, test file

T-04: Room-radar Return works after canceled dialog (R-RM-01)

Priority: high · Tier: S · Linked plan: Phase 0A

What to do: Return sets busyRef=true then only opens dialog; Cancel never runs returnMut.onSettled → later taps blocked. Reset busyRef on dialog close via onOpenChange.

Acceptance criteria:

  • RED then GREEN: tests/room-radar-return-busyref.test.tsx
  • pnpm test -- tests/room-radar-return-busyref.test.tsx && pnpm typecheck
  • Proof logged

Files in scope: src/pages/room-radar.tsx, test file

Owner / ops — Phase 0B + exit (not agent TDD)

Tier: Owner. Binary checks — see plan § Phase 0B. Not RED→GREEN.

ID Summary Status
T-06 Rostered reviewer account + active shift (highest-value) ready (Owner)
T-07 Build only via pnpm cap:build:native ready (Owner)
T-08 SIWA round-trip if social login retained ready (Owner) / N/A if email-only
T-09 Sentry ↔ PrivacyInfo / ASC privacy ready (Owner)
T-10 Broaden NSCameraUsageDescription ready (Owner)
T-11 Localize permission prompts (he) ready (Owner)
T-12 Offline cold-start "connect to sign in" ready (Owner)
T-13 AASA + entitlements live ready (Owner)
T-14 auth:preflight + validate:prod + verify:resubmission ready (Owner)
T-15 App Review notes framing ready (Owner)
T-16 Phase 0 exit on-device drill (blocks Phase 1) historical — void 2026-08-13; survives only as store-lane O8 (G3 on-device verdict) in the store-submission runbook

Queued — Phase 1+ (⚠ the "do not start until T-16 passes" gate is VOID as of 2026-08-13)

These cards were written in July. Much of this list either shipped under a different name (equipment fixes, shift/home, web admin-gate, Code Blue races, board features) or was superseded by the RN migration. Re-verify against git log before executing any card here.

Full cards in the plan library. Summary only:

Bundle IDs Plan
Equipment fixes T-17…T-21 phase-0-1.plan.md
Locate / readiness / damaged T-22a–c, T-23a–e, T-24a–e same
Shift / home T-25…T-27; R-SH-F1 sub-spec same + subspecs/R-SH-F1-shift-handover.plan.md
Inventory T-28a–b, T-29, T-30a1…T-30c phase-0-1.plan.md
Web admin-gate T-31 (S +R) same
Code Blue races (gate medium-01) R-CB-02/03 subspecs/R-CB-stabilize-code-blue-races.plan.md
Phase 2 MED + features T-34…T-44; R-CBF-1 / R-BDF-1 / R-PDF-1 phase-2-3.plan.md + sub-specs
Phase 3 LOW T-45…T-53 phase-2-3.plan.md
RFID-gate e2e R-M1.0…M1.5 subspecs/R-M1-rfid-gate-e2e.plan.md
Phase 4 parked massive-03, medium-04 phase-4.plan.md — entry conditions required

Blocked

  • T-16 / Phase 1+ — blocked until Phase 0A HIGH fixes + Owner 0B checks needed for the exit drill are done Historical — no longer blocked (void 2026-08-13). Phase 0A landed 2026-07-12; Phase 1+ either shipped or was superseded by the RN migration. The only live on-device gate is store-lane O8.
  • Phase 4 massive-03 / medium-04 — owner entry conditions (see phase-4.plan.md)

Completed

  • 2026-08-13 — The nine-blocker push + TV board (PRs #167–#181 to main): native push + RFID readers, reviewer-demo seed (#175), Code Blue vet-QA persona + click-path runbook (#179), doctor shift gate + migrations 181–184 (#180, RN companion #59), TV Command Center board 10-foot mode (#178) and the state-driven Phase-1 redesign (#181). All CI-green and proof-logged; deployed to production the same day (three Railway releases).
  • 2026-07-12 — Phase 0A: T-05, T-01, T-02, T-03, T-04 (HIGH fixes; RED→GREEN, batch gate green) — PROOF_ALIGNMENT_LOG entries "2026-07-12 — Consolidated Audit × 10x".

Archive completed tasks here with date and notes.

Backlog

Agents: add out-of-scope items here rather than acting on them.

Deferred from prior maintenance PLAN (superseded by Audit × 10x)

  • TASK-001: Eliminate N+1 queries on equipment list endpoint — was Area 2; reassess after Phase 0–1
  • TASK-002: Add missing test coverage for restock service (tests/restock.service.unit.test.ts)
  • TASK-003: Hebrew translation parity sweep (use pnpm i18n:check / parity tests when touching locales)

Ongoing

  • TASK: Investigate stale check-in sweep worker — confirm TTL sweep is running in production
  • TASK: Audit vt_event_outbox retention — verify janitor is not letting the table grow unbounded
  • TASK: Review Playwright Phase 9 drills — confirm all 8 drills pass against local dev server
  • TASK: Add .cursor/rules/ vettrack-specific overrides for i18n and multi-tenancy invariants
  • DONE (2026-07-07 relevance cleanup): deleted root cruft (Archive.zip, Archive 2.zip, all-files.md, screenshot.png, app-cloud.js, .nvrmc, session .txt, 38 playwright-ui-screenshots/); removed verified-dead shared/permissions.ts (server uses er-mode-permissions.ts), server/integrations/{rollout,conflicts}/*, src/lib/constants/regex.ts, src/lib/task-dashboard-filters.ts, src/hooks/use-is-mobile.ts. Gate green; superseded stale PR #40.
  • TASK: Remove dead src/features/today/* unused siblings (QuickScanCard, ShiftHero, TodayScreen, UrgentCountChips, use-today-shift, index.ts) — verify each is unreachable first. KEEP HomeTabletDashboard.tsx — it is LIVE (src/pages/home.tsx:26 imports it). PR #40's blanket features/today deletion was unsafe for this reason.
  • TASK: Audit docs/design-handoff/ (240 tracked files, ~15MB) — archive externally or trim to active design refs
  • TASK: NFC FAB for the native shell — never landed; reference implementation preserved at local ref refs/removed-origin/claude-new-session-rw4978 (origin branch deleted 2026-07-22 repo tidy). Built pre-src/app/platform seam — re-implement against the current platform router, do not rebase.
  • TASK: Deduplicate untracked .agents/skills/ecc/ mirror of .claude/skills/ecc/ (~1.7MB each); pick one canonical agent-skills path
  • NOTE (do not "clean"): inventory-deduction worker/queue is NOT dead — server/services/dispense.service.ts:614 enqueues it and 5 tests assert its shape. Removing it is a behavior change (Removal Protocol), not a relevance-cleanup deletion. Reassess only if the dispense enqueue path is intentionally retired.
  • TASK: Split remaining >800-LOC files (repo ceiling). admin.tsx DONE (1656→219 + prop-less section files under src/pages/admin/). Next, modular/clean: equipment-list.tsx (1351 — extract EquipmentItem + desktop sub-sections). Monolithic/higher-risk (single cohesive component; needs real decomposition + visual-regression, do individually): equipment-detail.tsx (2075), Tasks.tsx (1590), inventory-page.tsx (1033), route files containers.ts/users.ts/equipment.ts. EXCLUDE frozen/generated: i18n.generated.d.ts (generated), metrics.ts, code-blue.ts, realtime.ts, auth.ts, hand-built i18n.ts.

Task Template

### T-NNN / TASK-NNN: [Title]
**Priority:** `high` / `medium` / `low`
**Tier:** `S` / `S +R` / `O +R` / `Owner` (from plan card)
**Linked plan:** [phase-0-1 / phase-2-3 / subspec path + requirement id]

**What to do:**
[2–4 sentences. Prefer copying the plan card defect + GREEN direction.]

**Acceptance criteria:**
- [ ] RED then GREEN per plan card test file
- [ ] Card Verify command passes
- [ ] Proof logged in docs/audit/PROOF_ALIGNMENT_LOG.md
- [ ] No TODO comments in delivered code

**Files in scope:**
- (from plan card — ≤2 impl + 1 test unless mechanical mount fan-out)

**Files NOT in scope:**
- (from plan card guardrails)

**Notes:**
[Frozen doctrine, Tier review/drill gates]