Deploy Checklist Skill

Trigger

Load this skill when a dev is about to deploy or has just deployed:

• "We're deploying to production"
• "Pre-deploy check"
• "Post-deploy verification"
• "Run the deploy checklist"
• "Is it safe to deploy?"

Do NOT use this skill for: committing code (use git-os), creating a PR (use pr-create), or planning a project (use greenfield). This skill only applies at the deployment stage — code is already merged.

---

Run this checklist before and after every production deployment.

Pre-Deploy Checklist

• All CI checks green on the deploy branch
• Environment variables verified in target environment (no missing keys)
• Database migrations reviewed — irreversible migrations documented
• Migrations have been dry-run or tested in staging
• Rollback plan documented: what to revert and how
• Feature flags set correctly for the release
• Downstream services notified if API contracts changed
• Changelog updated with this release's changes
• Deployment window confirmed (avoid peak traffic)

Deploy

• Deploy initiated with correct branch / tag
• Deployment logs monitored in real time
• No unexpected errors during startup

Post-Deploy Checklist

• Smoke test: critical user flows verified manually or via synthetic monitoring
• Health check endpoint returns 200
• Error rate in monitoring (Datadog, Grafana, Sentry) is normal
• No spike in latency or DB query time
• Monitoring alerts reviewed — no new alerts triggered
• Changelog published / communicated to stakeholders
• Ticket status updated (closed / released)

Rollback Trigger Criteria

Initiate rollback immediately if:

• Error rate rises above 1% of requests
• P95 latency increases by more than 2x baseline
• Any data integrity issue detected
• Critical feature path returns 5xx

Tools

Action	Tool
Run lint, test, build scripts	Bash
Check health endpoint	Bash — curl -s <url>/health
Read config or env files	Read
Check CI status	Bash — gh run list or gh pr checks

Notes

• Never deploy on Fridays unless it's a critical hotfix
• Always have a second engineer available during production deploys
• Document the actual deploy time and outcome in the ticket