verify

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

• Third-party content exposure detected (high risk: 0.85). Yes: in SKILL.md Step 1 and Step 2b it reads outsider-authored GitHub issue bodies/comments at runtime via gh issue view ... --json body and gh api .../issues/<N>/comments ... .body, then quotes/uses that free text (including verbatim gate criteria) in the agent’s LLM context/output.