domains / families / agent-tooling / break-ai-fix-loops

break-ai-fix-loops

authentic

Stop ineffective AI coding repair loops with stable failure fingerprints, a three-attempt budget, real-path proof, negative controls, and tested rollback.

When to use

• Use when an AI coding agent cycles through similar patches without changing the observed failure.
• Use when a focused test passes but the installed, deployed, UI, API, persistence, or other real execution path still fails.
• Use when a repair claim needs a verifier that can reject a known-bad state and a rollback that has actually restored the baseline on a disposable copy.
• Do not use for a one-shot, already-understood edit whose acceptance check directly exercises the complete claimed behavior.

When not to use

• This workflow cannot prove a repair when the defect is not reproducible, the real execution path is inaccessible, or the primary verifier cannot observe the acceptance claim.
• A three-attempt budget exposes stagnation; it does not identify the correct architecture or replace domain expertise.
• A known-bad control demonstrates that one verifier catches one defect class. It does not prove complete test coverage.
• Rollback verification covers the recorded paths and baseline behavior only; external systems need their own provider-side readback.

Skill page

Source