mind-your-maxims

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill provides guidelines for the agent to critique its own communication style (e.g., avoiding 'buried ledes' or padding). It explicitly avoids safety-related topics by stating that 'Benevolence' is handled by the platform's safety layer, demonstrating a safe posture.
  • [EXTERNAL_DOWNLOADS]: The skill references established academic and informational resources including Wikipedia, University of Pennsylvania (UPenn), and arXiv. These are well-known, trusted domains used for providing context and documentation, not for automated downloads or execution of code.
  • [COMMAND_EXECUTION]: There is no evidence of shell command usage, dynamic code generation, or system-level calls in the SKILL.md or the evaluation tests.
  • [DATA_EXPOSURE]: The skill references a local context file using the ${CLAUDE_PLUGIN_ROOT} environment variable. This is a standard mechanism for local file referencing within the agent's plugin environment and does not involve exfiltrating or exposing sensitive user credentials or system files.
  • [DATA_EXFILTRATION]: No network tools or external data transmission patterns were detected. The skill operates entirely within the conversation context to evaluate its own previous outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:55 PM
Security Audit — agent-trust-hub — mind-your-maxims