feynman-perspective

Pass

Audited by Gen Agent Trust Hub on May 26, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill defines a cognitive framework and a specific persona based on historical and academic sources without attempting to override agent safety protocols.
  • [SAFE]: No prompt injection techniques or markers, such as instructions to ignore previous rules or bypass filters, were found in the text.
  • [NO_CODE]: The skill is entirely descriptive and does not include any executable scripts, binaries, or instructions for installing external software dependencies.
  • [EXTERNAL_DOWNLOADS]: While the instructions mention using tools like WebSearch to verify facts, this represents a standard use of agent capabilities for research and does not constitute a remote code execution risk.
  • [DATA_EXFILTRATION]: There are no patterns indicating attempts to access sensitive file paths, harvest credentials, or exfiltrate private user information to external servers.
Audit Metadata
Risk Level
SAFE
Analyzed
May 26, 2026, 01:01 AM
Security Audit — agent-trust-hub — feynman-perspective