skill-evolution

Pass

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: Uses curl to communicate with a local API service (localhost:18789) for pattern storage and search. These operations are limited to the local system and facilitate tracking research performance metadata.
  • [DATA_EXFILTRATION]: All network operations are performed against localhost, which is a whitelisted destination for local inter-process communication. No exfiltration to external domains or unauthorized servers is present.
  • [SAFE]: The skill establishes a framework for meta-analysis and self-improvement of agent skills using a local database (~/.openclaw/science-evolution.db). It does not include obfuscation, persistence, or privilege escalation mechanisms.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 6, 2026, 11:31 PM
Security Audit — agent-trust-hub — skill-evolution