watch-cli skill
watch-cli is a skill and CLI that hands an agent the frames and transcript of any social video so it can read the clip itself.
- 259 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Clip audio is sent to an external recognition service and `WATCH_BROWSER=auto` reads browser cookies.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/watch-cliDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Clip audio is sent to an external recognition service and `WATCH_BROWSER=auto` reads browser cookies.
This is third-party code. Review the repository files before installing.
What it does
watch-cli composes `yt-dlp`, `ffmpeg` and a speech-recognition model into one `watch <url>` command. Output is the video path, duration, evenly spaced frames and a transcript: the agent reads frames as images and speech as text. It works with YouTube, X, LinkedIn, TikTok, Reddit, Vimeo and Facebook. The repository ships five prompt templates that turn a video into a working project, an architecture diagram, a component, a notebook from a paper or a step-by-step cheat sheet. Claude Code gets a plugin and `SKILL.md`, other clients an MCP server from `mcp-server/`.
Who it is for. Developers and researchers who hand an agent a link to a video with code, a diagram or a talk and expect a working result from it.
Good fit when
- An agent must reproduce a project from a video walkthrough using frames and speech.
- An architecture diagram is needed from a talk where the diagram stays visible for at least three seconds.
- A long tutorial has to become a step-by-step cheat sheet.
- A video behind a login wall is read through your own browser cookies.
Not a fit when
- The video is fast-cut or music-driven: eight frames miss key moments.
- The clip runs past about two hours: the speech provider caps audio size.
- Audio cannot be sent to an external service.
Example request
Watch this video with watch-cli, describe the architecture from frames and transcript and build a diagram.Limitations
Speech recognition is paid and goes through Kyma API (the `KYMA_API_KEY` key) or your own Groq and Google keys. Audio is sent to an external service. For login-walled posts `WATCH_BROWSER=auto` reads cookies from a signed-in browser, so grant that access deliberately. Clips over two hours, live streams and region-locked videos work poorly. The repository's last commit dates from September 2026.
How to disable. Remove the plugin through `/plugin` or the `~/.claude/skills/watch-cli` folder, then the CLI the way it was installed, and clear the frame cache.
FAQ
What does the watch command output?
The video path, duration, a list of frames and a transcript. The agent reads frames as images and the transcript as text.
How much does transcription cost?
Only speech recognition is paid: about $0.005 for a five-minute clip and about $0.05 for an hour. Frame extraction runs locally.
How do I open a login-walled post?
Run `WATCH_BROWSER=auto watch <url>`: `yt-dlp` reads cookies from a signed-in browser. On servers without a browser pass a file with `--cookies`.
How do I get more frames?
Pass a second argument such as `watch <url> 24`. It helps on fast clips and scrolling code.
Is there an MCP server?
Yes, the `mcp-server/` folder holds an MCP server for Claude Desktop, Cursor and other clients. Claude Code only needs the plugin.
Discussion
No questions or reviews yet.
Related
SkillsThe /watch skill turns a link or file into frames and captions, or with a Gemini key hands the whole video to Google for the agent to answer from
A Claude Code skill that downloads a YouTube video, splits it into semantic chapters, cuts chosen clips, translates subtitles and burns bilingual ones in
Claude Code skill that searches YouTube, browses channels and downloads videos, audio and subtitles with yt-dlp.
Claude Video Vision
claude-video-vision
An MCP server and Claude Code plugin that let the agent watch video: frames via ffmpeg and audio transcription via Gemini, local Whisper or OpenAI
A skill and MCP server for cutting long videos: the agent takes a link or file and builds Shorts, Reels and TikTok clips with a cover, post text and ZIP
Claude Real Video
claude-real-video
A skill that lets Claude watch a video: it pulls scene-aware, deduplicated frames and a speech transcript through whisper from a link or file
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| watch-cli (this card) | no | 259 | |
| video-lens | no | 121 | another video reader for agents; watch-cli returns raw frames and transcript with no recap |
| yt-dlp | no | 205,517 | the downloader itself; watch-cli wires it to `ffmpeg` and speech recognition for an agent |
README in short
The README shows the output format, five prompt templates and install options, then covers cost and where the method works and where it does not.