Question
Your team runs a shared TypeScript monorepo. Three teammates each kicked off AI agents this morning to add unrelated features, and all three agents independently 'helpfully' touched shared utilities — one renamed a helper, one changed a function's signature, one added a dependency. Now you have three PRs that each work alone but conflict and break each other. How do you prevent this pattern going forward when everyone's using agents on the same repo?
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 coding: describe the solution in plain language (or narrate it) and the coach grades your approach. Generating runnable code from your description is coming next.