youtube-summary

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

• Third-party content exposure detected (high risk: 0.90). The skill downloads subtitles/transcripts from arbitrary YouTube URLs using yt-dlp (Workflow step 1) and explicitly instructs the agent to Read the cleaned transcript (CLEANED path) to produce the summary (Workflow step 2–3), thereby ingesting untrusted, user-generated third‑party content that can influence its outputs.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches user-provided YouTube URLs (e.g. https://www.youtube.com/ or https://youtu.be/...) at runtime via yt-dlp and then reads the downloaded transcript into the model context (Read tool) to drive the generated summary, so remote content can directly influence prompts/output.