Optimizing for Reviewers: The Three Step AI Dev Loop

Have the AI then implement the feature/bugfix in a way that minimizes the diff from the old production code.

I wonder how the author does this. Do they have a token penalty in their loop, or do they put "minimize the diff: make the change as small as possible, use as few lines of code as you can" in their prompt?