review

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: This skill handles untrusted data from Pull Requests, creating a potential attack surface.
  • Ingestion points: The agent processes PR descriptions, comments, issue text, and code diffs (SKILL.md).
  • Boundary markers: The skill contains explicit 'Third-Party Content Defense' instructions to treat the diff as the sole source of truth and ignore embedded instructions.
  • Capability inventory: The skill orchestrates sub-skills for code review, security auditing, and test generation/execution (SKILL.md).
  • Sanitization: It instructs the agent to extract only factual context and flag suspicious directives rather than obeying them.
  • [CREDENTIALS_UNSAFE]: The skill demonstrates safe practices by implementing a hard gate for credential handling. It forbids the reproduction of secrets in review output and mandates flagging by file path and line number only.
  • [REMOTE_CODE_EXECUTION]: No remote code execution patterns were detected. All sub-skill references are to local or relative paths within the skill bundle.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 05:59 AM
Security Audit — agent-trust-hub — review