This loop separates the builder from the reviewer so the agent does not grade its own work.
The cycle implements the next slice of the feature, hands it to an independent reviewer subagent, fixes every blocking finding, runs the tests, and reviews again until the review is clean.
Code keeps bouncing between builder and reviewer until it genuinely passes.
In the end you get a reviewed change with a green test suite.
When to use
- For code review without the self-grading bias.
- When you want a critical second pair of eyes before merge.
- On features where correctness and design really matter.
How to use
Run the loop asking it to implement the slice, hand it to an independent reviewer, fix the findings, and keep going until zero blockers with the suite passing.