claude-video
Agent skill that turns videos into timestamped frames and transcripts
What it is
A skill for Claude Code, Codex, and other Agent Skills hosts that lets a coding agent examine a video from a URL or local file. It downloads the video with yt-dlp, pulls timestamped frames with ffmpeg, and uses native captions or optional WhisperX, Groq, or OpenAI transcription. An optional Gemini mode sends the whole video to Google's model instead. Developers and researchers use it when they want an agent to summarize or answer questions about video content with timestamps.