Watch
An agent skill that gives a coding agent the ability to watch a video, turning a URL or local file into timestamped frames and a transcript.
Watch is an agent skill that gives an AI coding agent access to video as evidence, converting a URL or local video file into timestamped frames and a transcript the agent can read. Native captions are used first when available; when a video has no captions, optional local WhisperX transcription or a hosted Groq or OpenAI transcription service is used instead.
The skill installs into Claude Code through Claude Desktop's plugin marketplace, through the plugin commands in the Claude Code VS Code extension or terminal, or into Codex and other Agent Skills-compatible hosts by asking the agent to install it directly from the project's GitHub repository. It is documented as not working inside Claude Chat or Cowork, and browser-based agent apps are noted to have more limited support than the desktop and terminal integrations.
Once installed, a user starts a coding session and asks the agent to use the watch skill on a specific video to begin analyzing it.