Default output: return only the result, blockers, and required evidence. Omit preambles, process narration, repeated context, confidence scores, and follow-up offers. Use at most five bullets unless a required artifact or schema needs more.
Shipping and Launch
Overview
Shipping is review → PR → checklist → staged rollout → monitor. Small, frequent releases beat large batches. Do not skip review because launch checklists exist, and do not ship because staging looked fine without production safeguards.
When to Use
- Reviewing and opening/updating a PR before merge
- Preparing to deploy to production
- Launching a feature behind flags
- After code review and quality gates pass
End-to-end workflow
1. Review the change
Gather context: diff against base branch, uncommitted changes, recent commits, changed files, and user intent from recent chats if useful.
git fetch origin main
git diff origin/main...HEAD
git status
gh pr checks --json name,bucket,state,workflow,link
Then:
- Run targeted tests for changed behavior; add tests or document gaps.
- Review for correctness, regressions, security, and intent fit. Use parallel subagents on large diffs.
- Fix critical issues and re-run affected tests.
- Commit focused changes with a concise message.
- Push and open or update the PR.
Prioritize correctness, security, and regressions over style-only comments. Fix pre-commit failures; never bypass hooks. Use gh pr checks as the source of truth for PR readiness.
Output: findings (critical / warning / note), tests run, PR URL.
2. Complete pre-launch checklist
Code quality: tests pass, review approved, no critical/high issues, coverage meets threshold.
Security: no secrets in code/config, dependency audit clean, inputs validated, auth tested.
Performance: baseline established, no CWV/regression, no N+1, load tested if throughput-sensitive.
Operational: feature flag ready, monitoring configured, rollback tested, staging verified.
3. Deploy and roll out
- Deploy to staging; run smoke tests; check dashboards.
- Roll out in stages: internal → canary (1–5%) → expanded (25–50%) → GA.
- Monitor at each stage; rollback if metrics degrade.
- After GA, monitor 24–48 hours and schedule flag cleanup.
4. Rollback when needed
Automatic triggers: error rate, SLO breach, health check failures.
Manual rollback: identify bad deployment → revert to last good → verify → communicate → root-cause before redeploy.
Feature flag lifecycle
Create (off) → develop behind flag → test internally → gradual rollout → GA → remove flag and dead code path.
Anti-Rationalization
| Excuse | Counter | |--------|---------| | "It works on staging, ship it" | Staging does not replicate production traffic and data. | | "We can monitor after launch" | Monitoring must exist before launch. | | "Rollback is too slow, fix forward" | Fix-forward under pressure adds risk. | | "Staged rollout is too slow" | Staged rollouts beat incident response. | | "Flag cleanup can wait" | Flags become permanent debt. Schedule cleanup now. |
Verification
- [ ] Review complete; critical issues resolved
- [ ] PR checks green
- [ ] Pre-launch checklist complete
- [ ] Feature flag and monitoring configured
- [ ] Rollback tested
- [ ] Stakeholders informed