video-analysis

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly accepts and forwards public third-party video URLs (SKILL.md "公网可访问的视频 URL" and scripts/analyze_video.py which builds a message containing {"type":"video_url","video_url":...} and calls the GeekAI/OpenAI-compatible API) and also ingests douyin-video-fetch JSON outputs to extract public video links, so untrusted user-generated content is fetched/processed by the upstream model and can therefore influence outputs and subsequent agent behavior.