babysit-pr-to-pass-ci

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions targeting the agent's logic.
  • Ingestion points: The skill reads pull request diffs, GitHub Action job logs, and workflow files (SKILL.md).
  • Capability inventory: The agent has the authority to write to the filesystem, commit code, push to remote repositories, and rerun GitHub Actions jobs.
  • Sanitization: There are no explicit instructions for sanitizing or escaping the content of logs or diffs before they are processed by the LLM.
  • Boundary markers: The skill does not define strict delimiters (like XML tags or unique markers) to isolate untrusted log/diff content from the agent's primary instructions.
  • [COMMAND_EXECUTION]: The skill utilizes command-line tools to interact with the repository and CI environment.
  • Evidence: It executes gh run rerun to trigger GitHub Actions and uses git commands for committing and pushing code changes. These actions are performed as part of the intended babysitting functionality but represent a significant capability tier when combined with the ingestion of untrusted data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 03:22 PM
Security Audit — agent-trust-hub — babysit-pr-to-pass-ci