A multimodal AI agent skill that combines timestamped transcripts with visual frame extraction to provide comprehensive YouTube video analysis.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install youtube-knowledge-extractor
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install youtube-knowledge-extractor using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is YouTube Knowledge Extractor?
The YouTube Knowledge Extractor is a powerful addition to the Openclaw Skills library, designed to bridge the gap between what is said and what is shown in video content. While most tools rely solely on text-based transcripts, this skill performs a dual-channel analysis by extracting keyframes and synchronizing them with spoken dialogue. This allows for a much higher level of accuracy when documenting software tutorials, UI demonstrations, or complex explainer videos where visual context is critical for understanding the subject matter.
By leveraging tools like yt-dlp and ffmpeg, this skill enables AI agents to see screenshots, code blocks, and UI elements directly alongside the transcript. It is specifically engineered to handle technical hurdles like YouTube rate limiting and subtitle formatting, ensuring a robust workflow for developers and researchers who need to transform video content into structured, actionable documentation.
YouTube Knowledge Extractor Use Cases
- Generating step-by-step technical guides from software video tutorials
- Creating comprehensive summaries of educational videos with visual anchors
- Extracting code snippets, diagrams, or UI configurations shown on screen
- Analyzing long-form webinars to identify key visual transitions and spoken takeaways
How YouTube Knowledge Extractor Works
- Retrieve video metadata including duration and unique identifiers using yt-dlp.
- Access and parse internal YouTube json3 subtitle formats for precise timestamped segments.
- Extract visual frames using adaptive interval strategies or scene-change detection with ffmpeg.
- Map extracted images to their corresponding transcript segments based on timestamps.
- Perform a multimodal analysis to synthesize visual evidence with spoken context into a final report.
YouTube Knowledge Extractor Setup
To utilize this within the Openclaw Skills framework, ensure your environment has the necessary binaries installed. You can set up the required dependencies using the following commands:
# Install core dependencies
uv tool install yt-dlp
# Prepare local working directory
mkdir -p ./yt-analysis/frames
YouTube Knowledge Extractor Data Schema & Taxonomy
The skill organizes extracted data into a structured temporary directory to facilitate multimodal processing:
| Data Component | Format | Purpose |
|---|---|---|
| Metadata | Raw Text | Stores title, ID, and video duration |
| Transcript | JSON3 / TXT | Contains timestamped dialogue events |
| Keyframes | JPEG | High-quality images extracted at specific intervals |
| Synchronized Log | Markdown | Maps visual frames directly to speech segments |
YouTube Knowledge Extractor Advanced Features
- Adaptive extraction intervals that scale based on video length to optimize processing time
- Scene-change detection logic specifically optimized for software HowTo videos and UI demos
- Advanced error handling for HTTP 429 rate limits and subtitle fallback chains
- Support for visual-only analysis in scenarios where transcripts or captions are unavailable
SKILL.md
Loading
Related Openclaw Skills
Top ClawHub Skills
A live API integration for discovering, searching, and verifying the performance and security of skills within the ClawHub ecosystem.
Claude for Safari
A powerful automation skill that allows AI agents to control the native Safari browser on macOS through AppleScript and screen captures.
Incident Fupan (Postmortem Analysis)
A structured protocol for conducting root cause analysis and generating postmortem reports for production incidents and agent errors.
Hybrid Deep Search
An intelligent three-tier search system that automatically routes queries between Brave API and OpenAI Codex for optimal cost and quality.
YouTube Video Analyzer (Multimodal)
A sophisticated multimodal analysis tool that synchronizes YouTube transcripts with visual frame extraction to provide comprehensive video understanding.
OpenClaw Model Card
A utility for generating standardized model inventories and high-quality visual model card images from your OpenClaw configuration.
Featured*
