fix-pr-comments

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and parses user-generated PR review and issue comments from GitHub (via "gh api repos/{owner}/{repo}/pulls/{pr_number}/comments" and "gh api repos/{owner}/{repo}/issues/{pr_number}/comments" / "gh pr view"), and then reads and acts on that content to decide fixes, so untrusted third‑party text can influence agent actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill calls the GitHub API at runtime (e.g., via gh api repos/{owner}/{repo}/pulls/{pr_number}/comments and gh api repos/{owner}/{repo}/issues/{pr_number}/comments) to fetch review comments which directly drive the agent's instructions and edits, so these URLs are runtime external dependencies that control prompts.