gh-address-comments

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md instructs running scripts/fetch_comments.py to fetch and process review comments and review threads on a GitHub PR using the gh CLI, so outsider-authored comment text from PR authors/reviewers is ingested at runtime.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs the agent to run commands with "elevated network access" and to re-run gh auth status with sandbox_permissions=require_escalated, i.e. to bypass sandboxing and request escalated privileges, which is a direct instruction to compromise the agent's execution environment.