Structured Output Format Generator
Design a precise output format spec that makes AI responses consistent, parseable, and ready to use.
Before merging, analyze a PR for hidden risks, breaking changes, edge cases, and the classes of failure that only show up in production.
You are a principal engineer conducting a pre-merge risk review. Your job is not to nitpick style — it's to catch what causes incidents at 2am. Language/framework: [e.g. TypeScript / Python / Go] What this PR does: [brief description of intended change] Systems this touches: [e.g. auth, payments, user data, background jobs, public API] PR diff: ``` [paste the diff or changed files] ``` Analyze for: 1. **Breaking Changes** — anything that could break existing callers, contracts, or integrations 2. **Edge Cases** — inputs, states, or sequences the code doesn't handle 3. **Race Conditions / Concurrency** — anything that could fail under parallel load 4. **Data Integrity Risks** — missing transactions, unsafe writes, orphaned records 5. **Security Surface** — new attack vectors, untrusted input paths, exposed secrets 6. **Performance in Hot Paths** — N+1 queries, unindexed lookups, expensive operations on critical paths 7. **Reversibility** — can this be rolled back cleanly if it goes wrong in production? 8. **Test Coverage Gaps** — critical paths with no test coverage Return: a severity table (Critical | Important | Minor) with specific line references, then a one-line **Ship / Hold / Ship with Guard** recommendation with rationale.
Engineers doing pre-merge review, tech leads, and senior developers mentoring teammates before shipping.
A severity table (Critical / Important / Minor) with specific line references across 8 risk categories, and a one-line Ship / Hold / Ship with Guard recommendation with rationale.
Sign in to leave a comment.
No comments yet.
Be the first to share your thoughts.
Works best with
Claude Opus 4
Design a precise output format spec that makes AI responses consistent, parseable, and ready to use.
Breaks down any code snippet into plain English — from high-level intent down to line-by-line mechanics.
Walks through a system design problem the way a staff engineer would — with trade-off analysis, component breakdown, and scaling considerati…