youtube

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

• Third-party content exposure detected (high risk: 0.85). The skill calls YouTube Data API endpoints (e.g., search, playlistItems, commentThreads) and pipes the returned JSON fields (including comment textDisplay and other snippet text) into jq, which becomes readable LLM context; those texts are authored by non-operating-user parties (e.g., other creators/commenters), so this is outsider-authored free text via runtime API responses.