video-hosting-draft

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

• Third-party content exposure detected (high risk: 0.85). The skill ingests outsider-authored free text from transcripts/*.txt (or Whisper-generated transcript text from the raw .mp4 audio) and sends the first ~600 tokens to the LLM for category classification, so any transcript content originates from the production team/clip authors rather than the operating user.