Output includes an imperative-mood title (≤72 chars), a bullet summary of what changed, inferred motivation from the diff, and testing guidance.
It's mostly for the routine stuff like refactors and feature additions (diff tells most of the story). It won't help you explain a complex bug fix where the context is all in your head.
I would appreciate feedback, especially from those who have tried similar tools or think the Copilot overlap makes this redundant.