This is the classic loop: a builder writes code and a checker runs tests, typecheck, and lint, reporting exactly what broke.
The cycle implements the next item on the plan, runs the full check gate, feeds every failure back as the next instruction, and repeats until everything is green.
Where a one-shot agent ships its bugs, this loop catches them along the way.
It also adds a test for anything that lacked coverage.
When to use
- While building a feature, one plan item at a time.
- When you want a green build before anything ships.
- To close coverage gaps as they surface.
How to use
Run the loop asking it to build the next item, run tests, typecheck and lint, fix every failure, and stop only when the build is green.