| name | autonomy-recovery-loop |
| description | Autonomous failure recovery and self-healing loop for tasks, benchmarks, or services. Use when a run fails, dependencies break, or the system must diagnose, remediate, and re-run safely under guardrails. |
Autonomy Recovery Loop
Use this skill to recover from failed runs, missing dependencies, or unstable services.
Workflow
- Load a recovery plan (use assets/recovery_plan.json).
- Check dependency health and environment drift.
- Apply guardrails and select safe recovery actions.
- Run the minimal failing step.
- If it passes, re-run the full task.
Scripts
- Run: python scripts/recover.py --plan assets/recovery_plan.json --auto-start
Output Expectations
- Provide a short cause -> action -> result chain.
- State whether recovery succeeded and what fallback was used.