loop-fixer

To fix automated workflow plan failures, this Skill creates, executes, and verifies fix tasks based on evaluation reports. It manages iterative feedback loops between execution and evaluation steps to confirm issues are resolved.

An iterative feedback loop for debugging workflow execution automatically applies fixes to identified failures and then re-runs evaluations. This cycle repeats until the plan or execution successfully passes all critical evaluation checks.

To re-evaluate an automated plan after applying fixes, the fixer agent modifies the plan to address reported flaws and signals for it to be re-evaluated. This ensures the iterative improvement loop verifies that critical issues are resolved.

Yes, automating issue resolution requires access to the track's plan.md, spec.md, and metadata.json files. These files provide the necessary context for task management and state updates during the fix execution process.

The best way to handle critical flaws in an automated plan is to use an agent that parses evaluation reports, generates actionable fix tasks, and modifies the plan. It then signals for re-evaluation to ensure iterative improvement without manual intervention.

Workflow evaluation loops fail to resolve execution issues automatically if they lack an integrated fixer to create and verify fix tasks. Without parsing evaluation reports and managing iterative feedback loops between execution and evaluation, identified failures remain unaddressed.