github-review-pr

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires reviewers to include verbatim quotes of offending diff lines and to flag hardcoded secrets/credentials as findings, which forces the model to reproduce secret values in its output.

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

• Third-party content exposure detected (medium risk: 0.30). In the “Review GitHub Pull Request” skill, the required runtime workflow reads outsider-authored free text from a user-specified GitHub PR—via gh pr view (PR description, commit metadata, comments/reviews) and gh pr diff/gh api .../pulls/.../files (the diff, code comments, and commit messages)—before performing review angles and scoring.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches repository guidance at runtime (e.g., via gh api "repos/OWNER/REPO/contents/AGENTS.md?ref=BASE_SHA" and the templated gh api "repos/{REPO}/contents/PATH?ref={BASE_SHA}" / "repos/{REPO}/contents/PATH?ref={HEAD_SHA}"), and that fetched file content is injected into subagent prompts and used to direct review behavior, so it is a runtime external dependency that controls prompts.