big-review

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

• Third-party content exposure detected (high risk: 0.75). The required workflow can ingest outsider-authored free text via GitHub PR review mode: it fetches the PR description/body and commit messages (and full file contents) from the PR head ref using gh pr view / gh pr diff, then includes that text in the agent’s LLM context for the intent pass.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's GitHub PR workflow explicitly instructs fetching file contents at runtime via the GitHub API (e.g., "gh api repos/{owner}/{repo}/contents/{path}?ref={head} | base64 -d"), which would load remote files into the agent's context and therefore can directly control prompts/instructions used during the review.