Skip to content

Watch Videos

Lets the agent watch a video from a URL or file and answer questions about it

Stars
19,708
Licence
MIT
Files
12 files

What it does

Once installed, the agent can take a video link or a local video file, pull its captions or transcript, and extract still frames at timestamps so it can look at what is shown. You can then ask questions like what someone says or what happens at a given moment. If you set up a free Google Gemini key, Google's video model watches the full video, including audio, and answers directly instead.

Great for

  • People who want summaries or answers about a YouTube or other online video without watching it
  • Students or researchers pulling facts with timestamps from lectures or talks
  • Developers who want an agent to review screen recordings or demos

Skip it if

  • Sensitive or private footage, because with the Gemini engine YouTube URLs and uploaded local videos are sent to Google (local files are deleted after the answer), and cloud transcription sends audio to Groq or OpenAI
  • Visual questions when only the transcript is used, since the skill states transcript-only evidence cannot establish visual facts
  • Machines without about 3 GB free disk and 8 GB RAM if you want fully local WhisperX transcription, which is only verified on Apple Silicon macOS

Try asking

  • “Watch this YouTube video and summarize the main arguments with timestamps: <url>”
  • “What is shown on screen at 12:30 in this recording? <local file path>”
  • “Does the speaker in this talk mention pricing? Give the timestamp.”
  • “Extract the key steps demonstrated in this tutorial video”
Step-by-step guide: how to use Watch VideosInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 12 original files, the MIT licence and an attribution note.

    watch.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip watch.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip watch.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Choose the local engine if you do not want video or audio sent to Google, and check the requirements first
  • Ask for timestamps in your question so the agent can point you to the exact moment
  • Keep API keys in the config file at ~/.config/watch/.env rather than pasting them into commands

What's inside

  • scripts/build-skill.sh
  • scripts/config.py
  • scripts/download.py
  • scripts/frames.py
  • scripts/gemini.py
  • scripts/local_whisperx.py
  • scripts/runtime.py
  • scripts/setup.py
  • scripts/transcribe.py
  • scripts/watch.py
  • scripts/whisper.py
  • SKILL.md

Redistributed unmodified from bradautomates/claude-video under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.