review-prs

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

• Third-party content exposure detected (high risk: 0.90). The skill fetches PR diffs, prior review and issue comments from GitHub (see fetch-prs.py and prepare-review.py which embed "Here is the PR diff:" and prior comments into subagent prompt files), i.e., untrusted user-generated content that subagents are required to read and use to produce violations/comments that drive later posting decisions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill runs GitHub CLI/API calls at runtime (e.g., gh pr diff / gh api repos///pulls/ and gh api repos///contents/?ref=<base_branch>, i.e. https://api.github.com/repos/{repo}/pulls/ and https://api.github.com/repos/{repo}/contents/?ref=<base_branch>) to fetch PR diffs and content which are injected into subagent prompt files and therefore directly control the model prompts.