security-review

Pass

Audited by Gen Agent Trust Hub on Jul 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local script scripts/dep-scan.cjs using node to perform dependency scanning. This is a legitimate function for a security analysis tool.
  • [PROMPT_INJECTION]: The skill processes external data (repository code, PR diffs) which serves as a surface for indirect prompt injection. Malicious instructions embedded in the code being audited could potentially attempt to influence the agent's findings. * Ingestion points: Repository source code, pull request diffs, and default branch commits. * Boundary markers: None specified; the skill interprets raw code and diff data. * Capability inventory: The skill can execute local scripts and write results to the .bug-hunter/ directory. * Sanitization: No sanitization or input validation mechanisms are described for the analyzed content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 1, 2026, 08:09 AM
Security Audit — agent-trust-hub — security-review