This connects Claude to ByteDance's Volcano Engine video understanding API, specifically their doubao-seed models. It handles video uploads up to 512MB through the Files API, then analyzes content for scenes, actions, emotions, and answers questions about what's happening in the video. The recommended workflow is upload via Files API then analyze via Responses API, with automatic FPS sampling and 7-day file storage. You'll need an ARK_API_KEY to use it. It's seeing solid adoption with nearly 400 installs and has passed security audits across three platforms. Good fit if you're already in the ByteDance ecosystem or need video analysis capabilities beyond what standard vision models offer.
npx skills add https://github.com/freestylefly/canghe-skills --skill volcengine-video-understanding