symptomsync-review

# SymptomSync Review Use this skill when the task is review-oriented rather than feature implementation. ## Review priorities 1. Data-contract correctness 2. User-visible regression risk 3. Build, lint, or deploy breakage 4. Security and secret-handling issues 5. Drift between code, pipelines, and operator docs ## Repo-specific failure modes to check first - SQL changed in `supabase/` but not `database/`, or vice versa. - Frontend fields changed without updating `web/lib/*` schemas and helper payloads. - Review relies on `npm test` even though it is not authoritative here. - AI-service claims are taken from README instead of verified code. - Deployment changes ignore blue/green, canaries, `/health`, or SSM stage switching. - Docs claim behavior that the code no longer implements. Read `references/review-hotspots.md` for the most common review traps. ## Suggested review workflow 1. Classify the changed surface: web, data, agentic AI, release, or docs. 2. Read the nearest `AGENTS.md`. 3. Use the matching specialized skill when the diff spans a complex area. 4. Prefer concrete findings with affected files and downstream blast radius. ## Related skills - Use `$symptomsync-web-change`, `$symptomsync-data-change`, `$symptomsync-agentic-ai-change`, or `$symptomsync-release-change` for deep area-specific context.