skills/richardwu/agent-skills/fix-pr/Gen Agent Trust Hub

fix-pr

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It fetches review comments and PR comments using the GitHub API (gh api) and processes this untrusted text to generate code fixes. An attacker with commenting permissions could provide malicious instructions that the AI might execute. \n
  • Ingestion points: GitHub API calls for reviews and comments in SKILL.md. \n
  • Boundary markers: None present to distinguish between instructions and data. \n
  • Capability inventory: File reading/writing, shell command execution (bun, gh, git), and repository pushing. \n
  • Sanitization: None. \n- [COMMAND_EXECUTION]: The skill executes multiple shell commands (bun run test, bun run build, bunx biome) to verify fixes. Since these commands are run on code modified based on untrusted external comments, it creates a vector for executing malicious logic if the AI is successfully injected.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 04:35 PM