Ship it
Use when deploying a project to production - first deploy or an update - or when the user says ship it, deploy, go live, push to prod, or release.From its SKILL.md
npx -y skills add T4LEL/Claude-Arsenal --skill ship-itAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
- runs commandsInstructs the agent to run 7 commands, including `git status` and 6 more.
SKILL.md
2.9 KB, 648 tokens by cl100k_base, as published. Nobody here has run it
Ship It
Pre-flight → deploy → verify. Never skip pre-flight to "just get it out."
Copy this checklist and check off items as you complete them:
Ship-It Progress:
- [ ] Step 1: Pre-flight — every gate passes
- [ ] Step 2: Deploy (DB before code)
- [ ] Step 3: Verify live (mandatory)
- [ ] Step 4: Report
Step 1 — Pre-flight (every gate must pass; paste real output)
git statusclean or only intended changes; on the expected branch.- Tests pass. Typecheck/lint passes. Production build passes (
npm run build/ equivalent). If no test suite exists, record that as a risk in the report and run whatever checks do exist — never report a gate as passed that doesn't exist. - Env vars: diff
.env.exampleagainst the platform's configured vars (vercel env ls) — every required var present in production, no secret committed anywhere (checkgit ls-filesfor tracked env files). - Database: pending migrations identified; migrations are backward-compatible with currently-running code (deploy DB first, then code). Supabase: run the supabase MCP
get_advisorstool and triage security findings — RLS gaps block the deploy. - First deploy only: run the security-auditor agent on the codebase; Critical findings block.
A failed gate stops the deploy: fix it, or get the user's explicit go-ahead to ship anyway and record that in the report. Never reinterpret a failing gate as passing.
Step 2 — Deploy
- Next.js:
vercel --prod(or push to main if CI-driven — check which mode the project uses first). If the CLI isn't logged in or the project isn't linked, stop and ask the user to runvercel login/vercel link— never guess credentials. - Supabase migrations:
supabase db push(or the supabase MCPapply_migrationtool) BEFORE the code deploy. - Other stacks: follow the project CLAUDE.md's deploy section; if none exists, stop and write one first with the devops-engineer agent.
Deploying is outward-facing: on a FIRST production deploy or anything user-visible and irreversible, confirm with the user before executing unless they already said to ship.
Step 3 — Verify live (mandatory)
- Request the production URL — expect 200 and real content, not just exit code 0 from the CLI.
- Exercise the critical path once (signup/login/core action) via curl or playwright.
- Check platform logs for fresh errors (
vercel logs, supabase MCPget_logs). - If broken: roll back FIRST (
vercel rollback/ redeploy previous), debug second. Lead the report with the rollback.
Step 4 — Report
URL, what shipped, verification evidence, and any follow-ups (warnings triaged as non-blocking, advisors to revisit, gates the user explicitly waived).
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.