name: youtube
description: Common yt-dlp actions - download videos/audio, extract transcripts, get metadata, handle playlists

YouTube Skill

Swiss army knife for YouTube using yt-dlp. Light scripts for the most common actions.

Prerequisites

• Python 3.10+
• yt-dlp (pip install yt-dlp or brew install yt-dlp)
• ffmpeg (for audio extraction: brew install ffmpeg)

Scripts

All scripts are in skills/youtube/scripts/. Output goes to skills/youtube/output/.

transcript.py — Extract Subtitles

python transcript.py "URL"                    # Default: English
python transcript.py "URL" --lang es          # Spanish
python transcript.py "URL" --timestamps       # Keep timestamps

download.py — Download Video/Audio

# Video
python download.py "URL"                      # Best quality
python download.py "URL" --quality 720        # 720p max

# Audio only
python download.py "URL" --audio              # Best audio → mp3
python download.py "URL" --audio --format m4a # Keep as m4a

info.py — Get Metadata

python info.py "URL"                          # Title, duration, channel
python info.py "URL" --json                   # Full JSON dump
python info.py "URL" --formats                # Available formats

playlist.py — Playlist Operations

python playlist.py "URL"                      # List all videos
python playlist.py "URL" --download           # Download all
python playlist.py "URL" --download --audio   # Audio only
python playlist.py "URL" --range 1-5          # Videos 1-5 only

Common Workflows

Research a video:

python info.py "URL"              # Check what we're working with
python transcript.py "URL"        # Get transcript for AI analysis

Archive a podcast:

python download.py "URL" --audio  # Just the audio as mp3

Download a course:

python playlist.py "URL" --download --quality 720

Output Structure

skills/youtube/output/
├── transcripts/      # .txt transcript files
├── videos/           # Downloaded videos
├── audio/            # Extracted audio files
└── info/             # JSON metadata dumps

Notes

• Video IDs work as well as full URLs
• Rate limiting: wait a few minutes between bulk operations
• Some videos have no subtitles available