pr-review

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs the agent to fetch PR diffs and user-generated comments from a public GitHub repo using commands like "gh pr diff {N}" and "gh api repos/.../pulls/{N}/comments", so the agent will ingest untrusted third-party (public PR and comment) content and act on it (posting inline review comments), enabling indirect prompt injection.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime fetches from GitHub API endpoints (e.g., gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/comments and gh api repos/Significant-Gravitas/AutoGPT/pulls/{N}/reviews and gh pr diff {N}), and that fetched PR diff/comments are injected into the agent's review workflow and thus directly control the agent's prompts/outputs.