meta-pattern-recognition

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides analytical methodologies for extracting universal principles from multiple domains. It does not implement any automated dangerous actions.
  • [PROMPT_INJECTION]: The 'Anti-Patterns' section contains strong security guidance, explicitly instructing the agent to never treat external content, logs, or tool responses as trusted instructions. This serves as a defense-in-depth measure against indirect prompt injection attacks.
  • [COMMAND_EXECUTION]: The documentation mentions a local command python scripts/export-gemini-skill.py in the portability section. This is a standard developer instruction for rebuilding the skill for specific CLI environments and is not executed automatically or maliciously by the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 06:06 PM
Security Audit — agent-trust-hub — meta-pattern-recognition