review-github-pr

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

• Third-party content exposure detected (medium risk: 0.30). At runtime the workflow uses gh pr view <number> --json title,body,author,... and gh pr diff <number> then passes the resulting PR description and full diff as free text to sub-agents for analysis, meaning outsider-authored PR content is ingested directly.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill clones a user-supplied GitHub PR URL at runtime (e.g., "https://github.com/owner/repo/pull/123") and then reads and executes the repository's declared validation command from its CLAUDE.md, meaning remote repo content can control what commands are run locally.