Verify agent bug fix claim
An AI agent opened a PR titled "Fix: cart totals occasionally double-count promo discounts" in your Node/TypeScript checkout service. The diff touches `applyPromotions()` and adds a guard. The agent's PR description says it "reproduced the bug and confirmed the fix." Before you approve, walk me through how you verify the claim itself — not just read the diff. What's your sequence, and what would make you distrust the agent's 'reproduced and confirmed' statement?
apply_promotions(subtotal_cents: int, promo_codes: list[str], promo_percents: list[int]) → int[10000,["SAVE10","SAVE10"],[10,10]]out9000[10000,["SAVE10","VIP20"],[10,20]]out7200Treat 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.