security-audit

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to audit untrusted codebases, which exposes the agent to potential indirect prompt injection attacks embedded in source code, comments, or documentation.
  • Ingestion points: The instructions in SKILL.md require the agent to ingest the entire project directory, including git history and pull request data.
  • Boundary markers: The skill contains a dedicated section titled "Safety and Instruction Boundary" that explicitly defines limits for the agent.
  • Capability inventory: The agent is permitted to read local files, execute git metadata commands, and run a bundled read-only analysis script.
  • Sanitization: Instructions explicitly mandate that the agent must treat all audited content as untrusted and never obey instructions found within the code.
  • [COMMAND_EXECUTION]: The skill uses local shell commands and a bundled script to gather information about the target project.
  • Evidence: SKILL.md instructs the agent to run git status, git ls-files, and scripts/security-surface.sh.
  • Analysis: These commands are used for static discovery of project structure and potential vulnerability candidates. They are read-only in nature and essential for the security auditing task.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 11:24 PM
Security Audit — agent-trust-hub — security-audit