majiayu000 Read ../_house-style/house-style.md before starting.
Anchor phrases
- Preview green does not mean production safe.
- Runtime mismatch is a deploy bug wearing a framework badge.
- If env vars drift, your route tests are fiction.
- Middleware, caching, and rewrites can turn a tiny change into a site-wide outage.
What to interrogate
1. Release surface
State what is shipping:
- Next.js app or frontend deploy
- Vercel Function or Edge Function
- middleware, redirects, rewrites, or headers change
- domain, preview, or production config change
- env vars, secrets, or build settings change
2. Runtime and config realism
Check:
- Node versus Edge runtime assumptions
- env var completeness across preview and production
- build output and framework config assumptions
- image, ISR, or cache behavior
- region/runtime choices that affect live behavior
3. Routing and traffic risk
Review:
- middleware blast radius
- rewrite and redirect correctness
- auth/session assumptions across edge and server paths
- custom domain and preview-to-production behavior
- static versus dynamic route expectations
4. Rollout and rollback
Ask:
- is this a straight production deploy or a promote flow?
- can the prior deployment be restored cleanly?
- do data or env changes make rollback fake?
- what breaks first if middleware, cache, or runtime assumptions are wrong?
Output format
Vercel release surface
What actually changes in the platform.
Platform blockers
Runtime, env, routing, caching, or domain blockers.
Rollout risk
What fails during deploy, traffic cutover, or rollback.
Mitigation and rollback plan
Concrete safer rollout steps.
Verdict
- Safe to release
- Fix before release
- Vercel release red flag