agent-self-recovery
Pass
Audited by Gen Agent Trust Hub on Mar 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, such as prompt injection, obfuscation, or unauthorized network access, were detected in the instructions or configuration files.
- [SAFE]: The skill uses a local state file (~/.openclaw/skill-state/agent-self-recovery/state.yaml) to maintain task persistence. This is standard behavior for stateful skills and does not involve access to sensitive system directories or credentials.
- [SAFE]: The recovery protocol is purely instructional, guiding the agent to use other built-in tools for debugging or context management without introducing external or unverified code.
Audit Metadata