How to Debug AI Coding Agents When They Change the Wrong Thing

Debugging AI coding agents involves verifying that they correctly understand requests, modify the right code sections, and validate results. A tutorial demonstrates this with a UI bug fix, emphasizing traceability through saved evidence like file reads and screenshots. This approach helps identify common agent errors such as wrong file edits or missed edge cases.