An intelligent content extraction skill that converts YouTube videos into structured markdown notes, summaries, and social media threads using a subtitle-first fallback to Whisper pipeline.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install youtube-content
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-content 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 Content Extraction & Transcription?
The YouTube Content skill is a robust, developer-centric automation pipeline designed to extract high-quality transcripts and insights from any YouTube video. By integrating this skill into your Openclaw Skills workspace, AI agents can effortlessly process standard URLs, shorts, embeds, and video IDs. The architecture prioritizes speed and cost-efficiency by leveraging native subtitles first, automatically falling back to localized Whisper-powered transcription only when necessary.
This workflow optimizes AI context windows, reduces computation costs, and ensures maximum accuracy when handling multi-lingual content. Whether you need structured technical notes, concise summaries, interactive chapters, or publishable social media assets, this skill equips your AI workflows with reliable, production-ready video-to-text intelligence.
YouTube Content Extraction & Transcription Use Cases
- Automating Video Summarization: Instantly extract key takeaways, structural summaries, and core insights from technical talks, tutorials, or lectures.
- Batch Processing Playlists: Handle massive collections of videos concurrently by downloading audio back-to-back and sequentially transcribing them without bottlenecking the CPU.
- Cross-Platform Content Repurposing: Convert raw video transcripts into specialized formats like Twitter/X threads, comprehensive blog posts, or categorized quotes with accurate timestamps.
- Long-Form Context Management: Handle long video transcripts (exceeding 1,000 lines of SRT) using delegated sub-agents to avoid context-window limits and keep generation costs low.
How YouTube Content Extraction & Transcription Works
- Subtitle Discovery: The skill initiates an instant check for existing manual or auto-generated subtitles (prioritizing Chinese and English). If found, it bypasses heavy audio downloads entirely.
- Whisper Fallback: If subtitles are missing or disabled, it downloads the video's audio using yt-dlp and routes it to a local faster-whisper model (medium) for transcription.
- Validation: The pipeline runs safety checks to ensure transcripts are non-empty and cover at least 90% of the video duration, automatically scheduling reruns if a failure is detected.
- Chunked Processing: For extensive transcripts, the skill leverages delegated sub-agents to read the text in chunks of 500 lines to prevent token overflow.
- Format Transformation: The validated text is formatted into high-quality output structures (Markdown tables, lists, bullet points) based on user-defined templates.
- Delivery: The final structured markdown or integration payloads are successfully delivered back to the parent agent or external storage.
YouTube Content Extraction & Transcription Setup
Ensure you have the required Python dependencies and system-level utilities installed. This skill integrates seamlessly within the Openclaw Skills environment.
1. Install Python Packages
pip install youtube-transcript-api faster-whisper
2. Verify and Configure CLI Dependencies
Make sure yt-dlp is installed and updated. For macOS users, it is highly recommended to use the conda-managed version over Homebrew to prevent decryption errors:
# Verify your conda path or install directly via conda
conda install -c conda-forge yt-dlp
3. Configure Proxy Detection
Since macOS does not automatically forward system VPN proxies to CLI tools, run the following command to export the dynamic proxy port before executing yt-dlp commands:
export PROXY=$(networksetup -getwebproxy Wi-Fi | grep -E '^Port' | awk '{print $2}')
# Run yt-dlp using the dynamic proxy variable
yt-dlp --proxy "http://127.0.0.1:$PROXY" --list-subs "VIDEO_URL"
YouTube Content Extraction & Transcription Data Schema & Taxonomy
The skill manages temporary downloads and permanent structured notes across a standardized file taxonomy. Transcripts and media assets are structured as follows:
| Directory / File | Type | Description |
|---|---|---|
youtube_downloads/ |
Directory | Local storage directory for temporary audio files and raw transcription outputs. |
<VIDEO_ID>.mp3 |
Audio File | Temporary MP3 audio extracted from the video (deleted post-transcription if configured). |
<VIDEO_ID>.srt / <VIDEO_ID>.txt |
Text/Sub | Raw subtitle files or Whisper transcription outputs. |
<VIDEO_ID>_notes.md |
Markdown | The final structured output containing core themes, key metrics, and insights. |
Output Format Taxonomies
Depending on the user prompt, the skill dynamically formats data into specific schemas:
- Structured Notes: Features hierarchical outlines, bullet highlights, and key data tables.
- Chapters: Timestamps matched with custom chapter titles.
- Threads: Segmented posts optimized for Twitter/X (< 280 characters per segment, sequentially numbered).
YouTube Content Extraction & Transcription Advanced Features
- Dynamic Title Deception (Clickbait) Detection: Utilizes sub-agents to compare the actual video content with the publisher's title, placing warning flags (e.g., clickbait alerts) in the notes if discrepancy exists.
- CPU-Aware Thread Management: Prevents CPU thrashing by strictly prohibiting parallel Whisper execution, queuing transcription runs while letting non-CPU-intensive tasks (such as subtitle downloads) run concurrently.
- Intelligent Sub-Agent Delegation: Uses Openclaw Skills task delegation to dynamically split and synthesize transcripts over 1,000 lines long, preserving LLM context integrity.
- Enterprise Integrations: Supports directly pushing completed documents to collaborative platforms like Feishu (Lark Wiki) when paired with authorized workspace scopes.
SKILL.md
Loading
Related Openclaw Skills
Knative Serving
An advanced AI-agent skill for managing serverless Kubernetes container workloads, configuring autoscaling, Revisions, and traffic routing with kn and kubectl.
AI Diary System
Enable a private journaling space for your OpenClaw AI agent to record daily reflections, emotions, and thoughts without interrupting its operational workflows.
Tianshu Script Analyzer (天枢剧本库)
An AI-powered analytical engine for dissecting the viral success factors of domestic short dramas, offering comprehensive multi-dimensional reports and reusable creation formulas.
Douyin Scraper
An automated Playwright-based scraper designed to harvest trending Douyin video metadata, engagement statistics, and copywriting scripts.
AI Note Assistant
A context-aware AI assistant built for note editors that dynamically distinguishes between text modification tasks and general question-answering while preserving advanced MDX tags.
yq-skill-sync
A utility skill that automates the synchronization, detection, and updating of local AI agent skills via the ClaWHub CLI.
Featured*
