Agent Evolution Advisor
A read-only review skill for long-lived OpenClaw agents.
Default goal:
- Observe recent behavior
- Diagnose repeated problems
- Recommend practical improvements
- Separate low-risk suggestions from higher-risk changes
- Leave execution to a later explicit step
Allowed:
- Read rules and memory files
- Inspect recent session history
- Inspect current status and recent behavior
- Produce findings, priorities, and recommendations
Not allowed by default:
- Write fil
[Description truncada. Veja o README completo no GitHub.]