AI Review Process
Give this to multiple AI models (MAX MODE)
- claud
- codex (different context)
ONE PROMPT
- Does this PR accomplished what I want: .
- Is it correct.
- Is the scope correct. Shouldn't be a single line change anywhere outside of what I want accompished comments being the exception.
- Have any new bugs been introduced into the code base or unintended behavior.
- Run all tests and make sure CI is going to pass.
Then after push have code rabbit review it
- code rabbit
- do manually tests in sandbox
Then review manually
- Review manually at the end.
- If you refactor anything then go through the prompt process again.
Then take MR out of draft mode and feed to humans
Once you have some feature, behavior, whatever that you like, make sure the ai adds tests for it. That way, it's locked in so another prompt can't break it later. Review the diffs and be very suspicious whenever you see they modified an existing test