bootstrap-checks-from-prs

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/collect_prs.py utilizes the subprocess module to execute the GitHub CLI (gh) for retrieving repository and pull request metadata. This is a core function of the skill and utilizes well-known, legitimate tooling.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by converting external pull request data into instructions for the agent.
  • Ingestion points: The scripts/collect_prs.py script fetches data from pull request bodies, review comments, and issue comments.
  • Boundary markers: The generated markdown files (produced by scripts/generate_check_drafts.py) do not employ delimiters or specific system instructions to prevent the agent from following malicious commands embedded in the evidence or rationale sections.
  • Capability inventory: The skill uses scripts/write_checks.py to write files to the .agents/checks/ directory. Agents that consume these generated checks generally possess capabilities to read/write files and execute shell commands.
  • Sanitization: Sanitization is limited to basic whitespace normalization and YAML quote escaping, which does not prevent an attacker from embedding natural language instructions that the agent might inadvertently follow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 11:51 PM