video-watcher
by kartinw
Coding Agents & IDEsBrief
Analyze video content by extracting frames at regular intervals.
Description
What This Skill Does
Extracts frames from video files at configurable intervals using ffmpeg, then examines those frames to build an understanding of the video content. Works with MP4, MOV, AVI, MKV, and other common formats. Useful when you need to review or describe a video without playing it.
Frame-by-frame extraction lets you analyze video content without needing a media player or video streaming support in the agent environment.
When to Use It
- Reviewing a recorded meeting to find key moments
- Checking what scenes appear in an unlabeled video file
- Extracting UI screenshots from a screen recording
- Describing video content for accessibility or documentation
- Auditing footage for specific visual elements or text
Example Workflow
User asks: "What happens in this 3-minute tutorial video?"
- 1
Run extract_frames.sh on the video file at default 1 fps
- 2
Read printed metadata to get duration and resolution
- 3
Sample frames every 3-5 seconds given the video length
- 4
Examine each sampled frame for scene changes, text, and actions
- 5
Compile observations into a sequential description
A scene-by-scene summary of the video content with timestamps
Requirements
- ffmpeg installed on the system
Installation
clawhub install kartinw/video-watcherRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.