Single-task Agent Skills that fix what AI builders leave behind. Each one: when to use, numbered steps, real SQL and commands, pitfalls and a "done when" checklist. Plus a router skill that picks the right one, and a free quick scan to tell you where to start.
Then open your app folder and run: claude "Use ship-ready-start-here and check if this app is safe to launch"
The bundled ship-ready-quick-scan skill runs one standard-library Python script over your app folder: committed secrets, .env not git-ignored, secrets behind NEXT_PUBLIC_/VITE_, Supabase tables with RLS off (or switched off later), using (true) read policies, views that skip RLS, API routes with no caller check, unsigned Stripe webhooks. It reads files only: no writes, no network. Each FAIL names the free skill that fixes it.
In Claude Code: claude "Run ship-ready-quick-scan on this app" · Or directly: python3 skills/ship-ready-quick-scan/scripts/quickscan.py path/to/app
A first pass, not an audit: it does not check npm advisories or run your app. The Crew's 59-check scan goes further.
Inside a Claude Code session: /plugin install ship-ready-lite --marketplace getshipready/ship-ready-skills · Claude Code older than 2.1.275: claude plugin marketplace add getshipready/ship-ready-skills, then claude plugin install ship-ready-lite@ship-ready
Just one area? claude plugin install ship-ready-supabase --marketplace getshipready/ship-ready-skills · Cursor: copy skills/* to .cursor/skills/
secret-leak-checkFind API keys, service-role keys and tokens committed to an AI-built repo or shipped in the frontend bundle, and produce a rotation list.
leaked-key-rotationRotate leaked API keys fast without breaking production: triage every secret that sat in the leaked place, cap spend with no deploy, revoke and reissue at each provider (OpenAI, Anthropic, OpenRouter, Stripe, Supabase, GitHub), update host env vars, redeploy, prove the old key is dead and check usage and billing since the leak.
public-env-auditAudit NEXT_PUBLIC_, VITE_, EXPO_PUBLIC_ and REACT_APP_ variables to make sure no secret is compiled into the browser bundle.
xss-sanitizeFind and fix places where user-generated content is rendered as HTML (dangerouslySetInnerHTML, v-html, innerHTML, markdown with raw HTML).
security-headersAdd Content-Security-Policy, HSTS, frame-ancestors, nosniff and Referrer-Policy to a Next.js, Vite or static app on Vercel or Netlify without breaking Supabase, Stripe or analytics.
api-rate-limitProtect login, signup, AI/LLM, email/SMS and other costly endpoints with per-user or per-IP rate limits and input caps so nobody can burn your credits or spam your users.
dependency-auditFind and fix vulnerable or abandoned npm packages in production dependencies without breaking the app, including Next.js security patches.
ssrf-url-fetch-guardSecure features that fetch a user-supplied URL (link previews, import from URL, webhooks you call, AI browsing) against SSRF and internal network access.
rls-quick-checkFive-minute check of a Supabase project's Row Level Security: which tables have RLS off, which policies are 'using (true)', which updates lack WITH CHECK, and which buckets are public.
rls-policy-writerWrite correct Supabase RLS policies for common access models: owner-only, public-read/owner-write, admin, team/multi-tenant, and server-only tables, with WITH CHECK, explicit Data API grants (required for new tables from Oct 30, 2026), indexes and a rollback.
mass-delete-guardStop one user, bug or leaked key from wiping a Supabase/Postgres table, and recover fast if it already happened: incident steps, delete audit log, soft delete, owner-only delete policies and a delete rate limit.
storage-bucket-lockdownMake Supabase Storage buckets private where needed, add per-user folder policies, and switch the app to signed URLs.
service-role-splitRemove the Supabase service_role key from browser code and split into a browser client (anon key + session) and a server-only admin client.
schema-to-migrationsTurn a Supabase project built through Lovable, Bolt or the dashboard into versioned SQL migrations so changes are reviewable, reproducible and testable on a branch.
backups-and-restoreMake sure a Supabase/Postgres app can recover from deleted data or a bad migration: check the backup plan, add a logical export, and run a restore drill.
edge-function-authMake Supabase Edge Functions verify the caller's JWT, check roles, use secrets correctly and handle CORS, especially in Lovable/Bolt SPAs where Edge Functions are the only server code.
teams-multitenant-rlsAdd organizations or teams to a Supabase app with membership-based RLS, roles per team, invitations and safe switching between workspaces.
protect-routes-nextjsAdd real server-side protection to pages and API routes in a Next.js App Router app using @supabase/ssr middleware and getUser(), replacing client-only redirects.
auth-redirect-urlsFix Supabase/Clerk/Auth.js login, magic-link, OAuth and password-reset redirects that point at localhost or the wrong domain after deploying.
oauth-callback-fixImplement or repair the /auth/callback route that exchanges the code for a session with Supabase PKCE in Next.js, fixing 'auth session missing', loops and blank pages after Google/GitHub login.
email-deliverability-smtpMove Supabase auth emails off the default rate-limited sender to custom SMTP (Resend, Postmark, SES) with SPF, DKIM and DMARC, and brand the templates.
admin-role-setupImplement admin/staff roles safely: role stored server-side, users unable to promote themselves, admin checks on the server and in RLS.
session-bugsDiagnose 'logged out randomly', 'user is null on first load', hydration mismatches and stale sessions in Supabase/Next.js or SPA apps with a structured checklist.
stripe-webhook-verifyMake a Stripe webhook endpoint verify signatures with the raw body, handle retries idempotently and return fast.
stripe-checkout-serverCreate Stripe Checkout sessions on the server with fixed Price IDs, the logged-in user's id and correct success/cancel URLs, never trusting prices or user ids from the browser.
entitlements-tableModel paid access with an entitlements/subscriptions table written only by the Stripe webhook and read by the app and RLS, replacing client-set is_pro flags.
stripe-subscription-lifecycleHandle the full subscription lifecycle: trials, upgrades/downgrades, failed payments, cancellations and the Stripe Customer Portal.
stripe-go-liveSwitch a Stripe integration from test to live mode safely: live keys, live webhook endpoint and secret, live Price IDs, tax, receipts and a real-card test.
usage-credits-billingImplement prepaid credits or usage limits for AI features with an atomic ledger, so users can't overspend or race the counter.
env-var-inventoryBuild a complete inventory of environment variables across local, preview and production (Vercel, Netlify, Supabase secrets), fix missing ones and write .env.example.
vercel-build-fixDiagnose and fix Vercel/Netlify build failures in AI-generated Next.js or Vite apps: type errors, ESLint, missing env vars, Node version, case-sensitive imports, lockfile mismatches.
spa-hosting-rewritesFix 404s on refresh and deep links for Vite/React SPAs (Lovable, Bolt exports) on Vercel, Netlify, Cloudflare Pages or S3 by adding the right rewrites, plus caching headers.
custom-domain-setupConnect a custom domain to Vercel/Netlify with correct DNS, HTTPS, www/apex redirect and update every place that stores the old URL (auth, Stripe, OAuth, emails).
preview-vs-prodStop preview deployments from touching production data and money: separate Supabase projects or branches, Stripe test keys in preview, and protected previews.
rollback-planWrite and rehearse how to undo a bad deploy or migration in minutes: instant rollback on the host, reversible migrations, feature flags.
error-and-uptime-alertsAdd error tracking (Sentry), uptime monitoring and a health endpoint so you hear about outages before users tweet.
n-plus-one-fixFind and fix N+1 query patterns (a query inside a loop or per list item) in Supabase/Prisma apps using joins, embedded selects or batched IN queries.
index-advisorFind slow Postgres/Supabase queries and add the right indexes (foreign keys, RLS owner columns, sort columns) with EXPLAIN evidence.
bundle-dietShrink the JavaScript bundle of a Next.js or Vite app: find heavy dependencies (moment, lodash, icon packs, charts), lazy-load below-the-fold components and remove dead code.
image-optimizationFix slow images: next/image or responsive srcset, modern formats, correct sizes, lazy loading and priority for the hero, plus Supabase image transforms.
lighthouse-triageRun Lighthouse or PageSpeed Insights on key pages and turn the report into a short prioritized fix list for LCP, INP and CLS.
bug-repro-firstTurn a vague bug report ('it doesn't work') into exact reproduction steps, expected vs actual, logs and the smallest failing case before any code change.
ai-loop-breakerGet out of the loop where an AI builder keeps 'fixing' the same bug and breaking other things: freeze, diff, isolate, write a precise prompt with evidence, and checkpoint.
agent-delete-guardStop your own AI coding agent (Claude Code, Cursor, Codex) from deleting files, code or databases, and get things back when it already has: checkpoint before every session, deny rules, a read-only database role for the agent, and a recovery order.
typescript-errors-fixFix accumulated TypeScript errors in AI-generated code without `any` spam or ts-ignore: generate Supabase types, fix nullability, and re-enable type checking in builds.
add-regression-testsAdd a minimal, high-value test suite to a vibe-coded app: smoke E2E for signup/login/pay, API auth tests and RLS tests, running in CI.
refactor-giant-componentSafely split giant AI-generated components or route files into smaller pieces (data hooks, presentational components, utils) without changing behaviour.
export-and-run-locallyExport a Lovable, Bolt, v0 or Replit project to GitHub and run it locally with Cursor/Claude Code: env vars, Supabase connection, scripts, and a sanity build.
seo-metadataAdd titles, descriptions, Open Graph/Twitter images, sitemap.xml, robots.txt and canonical URLs to a Next.js or Vite app, and keep private app pages out of search.
legal-pages-draftDraft a privacy policy, terms of service and cookie notice that match what the app actually collects and which processors it uses (Supabase, Stripe, analytics, AI APIs).
analytics-and-funnelSet up privacy-friendly analytics and define 5–7 key events (visit, signup, activation, checkout start, paid) so you know where launch traffic drops.
launch-post-draftsDraft launch posts for X, Reddit, Product Hunt, Hacker News and Indie Hackers that follow each community's rules: hook, concrete proof, honest limits, one CTA.
transactional-email-setupSend product emails (welcome, receipts, notifications) from your own domain with Resend/Postmark, verified DNS, templates and unsubscribe handling.
Keep Lovable's scan on. It's a good first pass. Per Lovable's docs, its built-in scans do static analysis: they read your code and configuration.
Lite adds a second pass on your repo that runs locally, no code upload, and walks your agent through a two-account test: sign in as a second user and try to read, edit or delete the first user's rows, so you can see what the live app actually allows, not only what the config suggests. The same skills work on Bolt, Cursor, v0 or hand-written apps.
Lite skills fix one thing at a time. The Crew runs a 59-check scan, routes every finding to a named specialist, holds the gates and hands you a ship score. See the report it wrote for our demo app (49 findings, score 0/100).