PR description for AI-generated change
You used Claude Code to generate ~80% of a TypeScript change that migrates your service's date handling from a deprecated library to the native Temporal API, touching ~14 files. You've tested the happy path locally. Write the PR description. What do you state you verified, what you're explicitly unsure about, and how do you flag the AI-generated portions so reviewers calibrate their attention correctly?
rank_review_attention(paths: list[str], ai_lines: list[int], total_lines: list[int], traced: list[bool], top_n: int) → list[str][["src/dates.ts","src/api.ts","README.md"],[120,40,5],[200,60,5],[false,true,true],2]out["src/dates.ts","src/api.ts"][["a.ts"],[0],[10],[true],3]out[][[],[],[],[],3]out[]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.