launch-checklist

Execute a comprehensive pre-launch gate check against the canonical checklist, surfacing every blocking issue that must be resolved before the product goes live. Non-autonomous: presents a full findings report; nothing is auto-fixed — the user decides how to address each item.

Phases

1. Load checklist template — read .claude/templates/launch-checklist.md to get the canonical item list. If the template is missing, fall back to the built-in checklist defined in this skill. Check whether a previous run exists at docs/specs/launch-checklist.md and note any items already marked PASS.
2. Security: secrets check — run git log --all --full-history -- '.env' '.env.*' and grep the working tree for patterns matching real API keys (sk_live_, sk_test_, eyJ, -----BEGIN). Verify .gitignore covers .env, .env.local, and .env.*.local. Any real secret found in version control is an immediate BLOCK.
3. Security: RLS verification — read all migration files under supabase/migrations/ and all Supabase query code under lib/ and app/. Confirm that every tenant-scoped table has ENABLE ROW LEVEL SECURITY and at least one policy. Flag any table that is missing RLS as a BLOCK unless it is explicitly documented as public or service-role-only.
4. Billing: Stripe live-mode check — verify that the production environment's STRIPE_SECRET_KEY starts with sk_live_ (check .env.example notes or ask the user to confirm). Verify that the Stripe webhook endpoint is registered for the production URL with the correct event types. Verify that STRIPE_WEBHOOK_SECRET is set for production. Flag test-mode keys in production as a BLOCK.
5. Operational readiness — check for: (a) an error monitoring integration (Sentry, Axiom, or equivalent — look for SDK imports in lib/ or app/), (b) the existence of /privacy and /terms routes or static pages (legal stubs), (c) a working health-check endpoint or Vercel deployment health, and (d) the CI workflow in .github/workflows/ is present and passing. Mark each as PASS, WARNING, or BLOCK.
6. Security audit gate (BLOCKING) — read the ## Verdict section of docs/specs/security-audit.md. Apply the following logic: (a) if the file is missing → BLOCK and instruct the user to run /security-audit; (b) if the ## Verdict section reads BLOCK (open High/Critical findings) → BLOCK and instruct the user to resolve all High/Critical findings listed there before proceeding; (c) if the ## Verdict section reads PASS → this gate is satisfied.
7. Tests green gate (BLOCKING) — apply the following logic: (a) if no vitest.config.ts exists in the project root → BLOCK and instruct the user to run /test-setup to scaffold the test suite before launch; (b) if vitest.config.ts exists, run the project's npm test script (Vitest unit/integration) and, if a Playwright config exists, npx playwright test; if either command exits non-zero → BLOCK and instruct the user to fix failures and run /write-tests for any files lacking coverage.
8. Write report and verdict — fill docs/specs/launch-checklist.md with each item's status and a one-line note. Output a final verdict: PASS (all items PASS or WARNING only) or BLOCK (one or more BLOCK items remain, including the security audit gate or the tests green gate). Under BLOCK, list the blocking items in priority order with remediation steps — security audit failures and test suite failures must appear first. Show the report for approval before writing the file. Honor the review intensity: under full, walk through each checklist section for sign-off; under lean, get a single confirmation before writing the report; under solo, proceed and summarize after.

Output

docs/specs/launch-checklist.md + a PASS or BLOCK verdict with itemized findings. Writes nothing without approval.