mars

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill defines a persona for code auditing with no requests for excessive permissions or external network access.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes untrusted codebase files.
  • Ingestion points: Source code files processed via the path argument in the /mars command.
  • Boundary markers: No delimiters are defined to isolate untrusted code from agent instructions.
  • Capability inventory: The agent reads and analyzes local files.
  • Sanitization: No sanitization of code content is implemented.
  • [SAFE]: No obfuscated content, hardcoded secrets, or persistence mechanisms were detected.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 11:30 AM
Security Audit — agent-trust-hub — mars