gh-address-comments

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

• Third-party content exposure detected (high risk: 0.90). The script scripts/fetch_comments.py shells out to gh api graphql to fetch PR conversation comments, reviews, and review threads from GitHub (user-generated PR comments) and the skill instructs the agent to read/summarize/act on those comments, exposing it to untrusted third‑party content.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The prompt explicitly requests elevated network/sandbox permissions (e.g., "Run all gh commands with elevated network access" and "sandbox_permissions=require_escalated") which asks the agent to bypass or escalate sandbox/security controls, even though file edits themselves are project-specific.