loop-repair

What problem does it solve?

This Skill helps recover agent runs that have failed verification, avoiding endless thrashing by forcing a disciplined, evidence-based repair process.

Core Features & Use Cases

• Failure Classification: Distinguishes deterministic failures, rubric failures, flaky checks, regressions, spec gaps, and environmental blockers.
• Bounded Repair Workflow: Makes the smallest possible fix, reruns the same check, and records whether the change was productive.
• Escalation and Safety: Caps retry attempts, avoids scope creep, refuses test manipulation, and escalates to replan, revert, approval, or termination when needed.
• Use Case: A test suite turns red after an agent patch; this Skill identifies the likely failure mode, applies one targeted repair, and confirms whether the fix actually improved verification.

Quick Start

Use the loop-repair skill to inspect the failing verification output, classify the failure mode, apply the smallest bounded fix, and rerun the same check.