lipsync

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow uses user-supplied --input JSON containing video_url/audio_url, and the CLI then fetches and downloads from those (untrusted) URLs for the model pipeline, so outsider-authored free text could be ingested indirectly if the remote content includes instructions.