youtube-to-heygen-video

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly sends arbitrary public YouTube URLs to Gemini for direct analysis (see SKILL.md "Sends YouTube URL to Gemini 2.0 Flash" and scripts/youtube_to_heygen_video.py → extract_script_from_youtube using types.FileData(file_uri=youtube_url)), meaning untrusted, user-generated web content is ingested and interpreted by the agent and that extracted content is used to drive downstream actions (script generation and HeyGen video creation), enabling indirect prompt-injection from third-party video/transcript content.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill sends user-supplied YouTube URLs (e.g., https://www.youtube.com/watch?v=VIDEO_ID) to Gemini at runtime (file_uri passed into the model) and uses Gemini's generated text as the script fed to HeyGen, so remote YouTube content directly controls the agent's prompts/output.