youtube-analysis

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

• Third-party content exposure detected (high risk: 0.90). This skill fetches and ingests transcripts and metadata directly from public YouTube pages using youtube-transcript-api and yt-dlp (see scripts/fetch_transcript.py -> fetch_video / fetch_transcript_api / fetch_transcript_ytdlp) and then explicitly instructs the agent (see SKILL.md and scripts/analyze_video.py) to read and analyze that untrusted, user-generated transcript content, so third-party instructions in those transcripts can directly influence the agent's analysis and actions.