| name | loopify-debug |
| description | Use when a loop is stuck, repeating failures, flaking, patching the wrong target, hitting an authority conflict, or burning turns without convergence. |
loopify-debug
Rescue stuck loops.
When To Use
- The same failure repeats.
- Checks are flaky or invalid.
- The evaluator is weak.
- The loop is burning turns without convergence.
- The next useful step might be blocked or escalated, not another patch.
Do Not Use
- Do not keep applying cosmetic patches to repeated failures.
- Do not weaken checks just to escape a stuck loop.
- Do not ignore authority-boundary conflicts.
Workflow
- Read Loop Contract and trace.
- Classify the stuck pattern with
references/stuck-loop-taxonomy.md.
- Identify whether the issue is implementation, check, spec, context,
authority, or budget.
- Recommend recovery using
templates/recovery-plan.md.
- Write a blocked report if no authorized recovery exists.
- Patch only when recovery is clear and authorized.
Output
Diagnosis, recovery plan, or blocked/escalated report.
Validation
- The repeated pattern is named.
- The next diagnostic differs from the failed attempts.
- Blocked/escalated recommendations cite the contract rule.
Example
Use loopify-debug on docs/examples/stuck-loop-debug/trace.md and classify why the loop is not converging.
