Media Extract

Universal media extraction and analysis skill for Claude Code. Extracts structured knowledge from YouTube videos, web articles, local files, and pasted content.

Features

• YouTube videos - transcript-based analysis OR visual analysis via Gemini
• Web articles - fetch and analyze any URL
• Local media - video, audio, PDF, text files
• Pasted content - transcripts, meeting notes, any text
• Batch mode - multiple sources with comparison
• 11 workflows - summary, golden nuggets, chapters, quotes, workflows, commands, breakdown, full extraction, visual analysis, download, batch

Prerequisites

Required

• yt-dlp - YouTube data extraction
• jq - JSON processing

brew install yt-dlp jq

Required for visual analysis

Visual analysis (Gemini multimodal) lets Claude see what’s on screen in videos - code, terminal, UI, diagrams. It requires:

• Python 3.8+
• The gemini-api-dev Claude Code skill
• A Gemini API key

Install the Gemini skill (pick one method):

# Using Vercel Skills CLI
npx skills add google-gemini/gemini-skills --skill gemini-api-dev --global

# OR using Context7 Skills CLI
npx ctx7 skills install /google-gemini/gemini-skills gemini-api-dev

Then install the Python SDK (used by the extraction script):

pip install google-genai

Installation

1. Copy the skill to your Claude Code skills directory:

cp -r media-extract ~/.claude/skills/content/media-extract

2. (For visual analysis) Set up your Gemini API key:

cd ~/.claude/skills/content/media-extract
cp .env.example .env
# Edit .env and add your Gemini API key

3. Make the fetch script executable:

chmod +x ~/.claude/skills/content/media-extract/fetch.sh

4. Verify installation:

# Test yt-dlp
yt-dlp --version

# Test jq
jq --version

# Test Gemini skill + SDK
python3 -c "from google import genai; print('OK')"

Usage

The skill activates automatically when you share content with Claude Code:

# YouTube - transcript summary (default)
Summarize this: https://youtube.com/watch?v=...

# YouTube - visual analysis (sees code on screen)
Watch this and extract the code: https://youtube.com/watch?v=...

# YouTube - download
Download this video: https://youtube.com/watch?v=...

# Web article
Summarize this article: https://example.com/blog/post

# Local video
Analyze this video: /path/to/tutorial.mp4

# Pasted text
[paste transcript] Extract the golden nuggets from this

# Batch
Compare these two videos: [url1] [url2]

Available workflows

Configuration

On first use, a config file is created at ~/.media-extract/config.json:

{
  "save_transcripts": true,
  "transcript_dir": "~/.media-extract/cache/transcripts",
  "download_dir": "~/Downloads"
}

• save_transcripts - Cache YouTube transcripts for reuse (saves API calls)
• transcript_dir - Where cached transcripts are stored
• download_dir - Default directory for downloaded videos

Directory Structure

media-extract/
  SKILL.md              # Main skill file (router)
  fetch.sh              # YouTube data fetcher (yt-dlp wrapper)
  .env.example          # Gemini API key template
  scripts/
    extract.py          # Gemini visual analysis script
  templates/
    summary-template.md
    golden-nugget-format.md
    chapters-template.md
    quotes-template.md
    workflow-template.md
    commands-code-template.md
    breakdown-template.md
    multi-file-output-template.md
  references/
    workflows/          # Individual workflow instructions
      visual-analysis.md
      download.md
      summary.md
      golden-nuggets.md
      chapters.md
      quotes.md
      workflow-extraction.md
      commands-code.md
      breakdown.md
      full-extraction.md
      batch.md
      clean.md

Sub-commands

This skill also registers slash command variants:

• /media-extract - Default (auto-detect intent)
• /media-extract:visual - Force visual analysis mode
• /media-extract:batch - Force batch mode
• /media-extract:download - Force download mode

Troubleshooting

“yt-dlp not installed” - Run brew install yt-dlp

“jq not installed” - Run brew install jq

“google-genai package not installed” - Run pip install google-genai. Also make sure the gemini-api-dev skill is installed.

“GEMINI_API_KEY not set” - Copy .env.example to .env and add your key from Google AI Studio

No subtitles available - The video may not have auto-generated captions. Try visual analysis instead.

Video processing failed - The video may be too long for Gemini upload. Try a shorter clip or transcript-based analysis.