Every acceptance criterion on the ticket was met, and the bug shipped anyway. When an AI agent picks up your issues, the criteria are the only contract it has. Here’s how to write ones it can’t satisfy without doing the actual work.
A shared skill and a team plugin make everyone fast. That’s the problem this post is about: when AI writes the code and a human clicks approve, who actually understood it? Two review gates — an automated one and a human one — keep the developer, not the AI, accountable.
A developer pastes a connection string into ChatGPT. Another copies customer data to test something. We’ve all been close. Here’s how to think about what’s safe to share and what isn’t.