chore: automation audit 2026-07-06

.agents/architecture.md

@@ -631,6 +631,7 @@ src/
 │   │   ├── noise-registry.ts   # NoisePattern type + NOISE_REGISTRY — declarative noise filter definitions
 │   │   ├── client-filter.ts   # shouldDropClientEvent — iterates registry for client/both-scoped patterns
 │   │   ├── server-filter.ts   # shouldDropServerEvent — iterates registry for server/both-scoped patterns
+│   │   ├── replay-filter.ts   # isReplayHydrationError — filters replay recording events (beforeAddRecordingEvent)
 │   │   └── capture.ts         # captureSupabaseError, captureApiError
 │   ├── theme.tsx            # ThemeProvider + useTheme hook (light/dark/system, localStorage persistence)
 │   ├── toast.ts            # Lazy-loaded sonner toast wrapper to reduce initial bundle size
@@ -677,6 +678,6 @@ Root config files:
 
 ## Observability
 
-- **Sentry client**: session replay (10% normal, 100% on error), route transition tracking
+- **Sentry client**: session replay (10% normal, 100% on error), route transition tracking, replay hydration error filter (`beforeAddRecordingEvent` via `isReplayHydrationError`)
 - **Sentry server**: PII enabled, local variables, 10% trace sampling in production
 - **Health endpoint**: `GET /api/health` — checks DB connectivity via sequential HEAD pings (first warms connection, second measures warm latency), returns status, latency (last sample), per-sample breakdown, threshold, and dynamically computed region co-location info

metrics/weekly/2026-W28-audit.md

@@ -0,0 +1,102 @@
+# Automation Audit — Week 28
+
+## Summary
+- Automations active: 17
+- PRs opened by automations: 36
+- PRs merged: 36
+- CI fixes applied: 0
+- Bug issues created: 1 (#1464 — Sentry Replay hydration error regression)
+
+## Per-Automation Assessment
+
+### Feature Builder
+- Issues completed: 2 (#1452 Copy link action, #1453 Open in new tab)
+- PRs merged: 2 (#1457, #1460)
+- Success rate: 100%
+- CI fixes: 0
+- Recurring problems: none
+- Note: Both features shipped on June 29. No new features after that — backlog has been empty since.
+
+### Feature Planner / Product Improver
+- Issues created: 1 (#1454 — content preview snippets, flagged needs-human)
+- No new issues triaged. Backlog remains empty. All remaining open issues are `needs-human`.
+
+### Bug Fixer
+- Bugs fixed: 2 (#1459 suppressHydrationWarning, #1464 replay hydration filter)
+- PRs merged: 2 (#1462, #1465)
+- Success rate: 100%
+- CI fixes: 0
+- Recurring problems: none
+- Note: Both fixes were Sentry-related hydration issues. The second (#1464) was a regression of the first — the initial fix (#1462) addressed React hydration warnings but not Sentry Replay's independent DOM diffing. The Bug Fixer correctly identified the different interception point (`beforeAddRecordingEvent` vs `beforeSend`) on the second attempt.
+
+### PR Reviewer
+- PRs reviewed: 36
+- CI fixes: 0
+- False positives: none observed
+- All 36 PRs merged cleanly. No rework cycles.
+
+### Post-Merge Verifier
+- Verifications run: 4 (PRs #1457, #1458, #1460, #1462, #1465)
+- Regressions caught: 1 (#1464 — Sentry Replay hydration error persisted after #1462 fix, leading to #1465)
+
+### UI Verifier
+- Verifications run: 4 (PRs #1457, #1458, #1460, #1462, #1465)
+- Design violations found: 0
+
+### Incident Responder
+- Errors triaged: 0
+- Bug issues created: 0
+- Still blocked by Sentry MCP 401 (#1305, now 33 days old).
+
+### Performance Monitor
+- Published W27 performance report (`metrics/weekly/2026-W27-perf.md`).
+- No performance issues detected.
+
+### Tweet Drafter
+- Posts published: 21 (3× daily for June 29 – July 5)
+- PRs merged: 21
+- Functioning normally.
+
+### Daily Metrics
+- Snapshots committed: 7 (June 29 – July 5)
+- All merged successfully.
+
+### Weekly Recap
+- Published W27 recap (`metrics/weekly/2026-W27.md`).
+
+### Automation Auditor
+- Published W27 audit (`metrics/weekly/2026-W27-audit.md`).
+
+### Bug Fixer (additional)
+- 1 PR (#1458) documented the noModule polyfill exclusion from bundle budget — a documentation fix, not a code fix.
+
+### Stale Issue Reviewer / Needs-Human Requeue / PR Shepherd
+- No stalled `status:in-progress` issues detected.
+- No stuck PRs observed.
+- `needs-human` issues remain unaddressed (see Label Hygiene below).
+
+## Label Hygiene
+- Stalled in-progress issues (>24h): none
+- Backlog issues missing priority: none (backlog is empty)
+- needs-human issues open >7 days:
+  - #1216: feat: enable OAuth sign-in with GitHub and Google (39 days) — priority:3
+  - #1245: feat: add column calculation summary row (38 days) — priority:3
+  - #1252: feat: add row grouping to database table view (37 days) — priority:3
+  - #1305: bug: Sentry MCP integration returns 401 (33 days) — priority:1, blocks Incident Responder
+  - #1344: feat: add Storybook visual snapshot auto-update workflow (25 days) — priority:3
+  - #1454: enhancement: add content preview snippets to workspace home (7 days) — priority:3
+
+## Knowledge Base Updates
+- **AGENTS.md**: no changes
+- **architecture.md**: added `replay-filter.ts` to sentry directory listing; added replay hydration error filter mention to observability section
+- **conventions.md**: no changes
+- **design.md**: no changes
+- **quality.md**: no changes — all grades remain A, test counts (2,185 Vitest / 500 E2E) match actual
+
+## Recommendations
+
+1. **Address #1305 (Sentry MCP 401) — now 33 days old.** This is priority:1 and blocks the Incident Responder, Performance Monitor, and any Sentry-dependent automation. Production error monitoring has been manual-only for over a month. A human administrator needs to re-authorize the Sentry OAuth integration.
+
+2. **Triage or close the aging priority:3 needs-human issues.** Five issues (#1216, #1245, #1252, #1344, #1454) have been waiting 7–39 days for human input. Either approve them for the automation queue, provide implementation guidance, or close as not-wanted. The backlog has been empty for 3+ weeks — the Feature Builder and Bug Fixer are idle.
+
+3. **The hydration error fix cycle (#1459 → #1462 → #1464 → #1465) shows a pattern worth noting.** The first fix addressed the symptom (React hydration warning) but not the root cause (Sentry Replay DOM diffing). The Post-Merge Verifier correctly caught the regression, and the Bug Fixer correctly identified the different interception point on the second attempt. No process change needed — the system self-corrected within hours.