loop-fixer

To automate evaluation-fix loops, you can parse failure reports and translate them into concrete fix tasks. The loop-fixer executes these fixes, updates commit history, and re-checks until passing.

Converting an execution evaluation failure report into actionable fixes involves parsing the failure data and generating corresponding tasks in plan.md. This enables automated re-evaluation without manual intervention.

Yes, you can bound automated retry attempts for execution fixes using guardrails. The system applies bounded retries with clear handoffs back to the evaluator for re-checking until a PASS state is reached.

Plan and execution evaluation cycles handle commit history by executing fixes deterministically and updating commit history and metadata. This ensures the loop state is managed accurately throughout the bounded retry process.

Automated re-evaluation loops require manual intervention when failure reports are not translated into concrete fix tasks. Automating this translation eliminates manual steps and enables continuous re-checking until PASS.