An OpenClaw skill that archives YouTube playlists as markdown notes with AI-generated summaries, full transcripts, and auto-assigned tags.

• Archives YouTube playlists as individual markdown files
• Each video gets: metadata, full transcript, AI summary, and topic tags
• Two-phase pipeline: fast import (metadata only), then enrichment (AI calls)
• Works with any LLM provider (OpenAI, Gemini, Anthropic, Ollama, OpenRouter) or none at all
• Runs on macOS, Linux, and Windows

---
title: "How to Self-Host Everything"
channel: "TechChannel"
url: https://www.youtube.com/watch?v=abc123
video_id: abc123
published: 2026-01-15
tags: ["youtube-archive", "self-hosted", "tutorial"]
enriched: true
---

# How to Self-Host Everything

## Summary
## Key Takeaways
- ...

## Transcript
> [!note]- Full Transcript
> ...

• Python 3.7+
• yt-dlp (pip install yt-dlp)
• A browser signed into YouTube (for private playlists)

No other dependencies. All HTTP calls use Python's stdlib urllib.

clawhub install youtube-archiver

Or clone this repo into your OpenClaw skills directory.

Then ask your agent: "Archive my YouTube playlists" and it will walk you through setup.

The scripts work standalone too:

# Initialize config
python3 scripts/yt-import.py --output ~/YouTube-Archive --init

# Edit .config.json with your playlists and preferences

# Import videos (fast, metadata only)
python3 scripts/yt-import.py --output ~/YouTube-Archive

# Enrich with summaries + transcripts
python3 scripts/yt-enrich.py --output ~/YouTube-Archive --limit 10

All preferences live in .config.json in the output directory:

{
  "playlists": [
    {"id": "LL", "name": "Liked Videos"},
    {"id": "WL", "name": "Watch Later"}
  ],
  "browser": "chrome",
  "summary": {
    "provider": "openai",
    "model": "gpt-5-mini",
    "api_key_env": "OPENAI_API_KEY"
  },
  "tagging": {
    "provider": "gemini",
    "model": "gemini-3-flash",
    "api_key_env": "GEMINI_API_KEY"
  },
  "tags": [
    "tutorial", "programming", "devops", "ai", "productivity",
    "design", "hardware", "open-source", "career", "self-hosted"
  ]
}

Works without any API keys. Summaries are skipped, tags fall back to keyword matching.

1. Import pulls playlist metadata via yt-dlp using browser cookies for auth
2. Each video becomes a markdown file with YAML frontmatter in a playlist subfolder
3. Enrichment extracts transcripts (via yt-dlp subtitles), generates summaries, and assigns tags
4. Idempotent: skips already-archived videos by video_id, skips already-enriched notes
5. Lockfile prevents concurrent runs

• macOS: Terminal needs Full Disk Access to read browser cookies
• Windows: Cookie extraction can be flaky; exporting a cookies.txt file is more reliable
• Linux: Works on desktop; headless servers need cookies_file

MIT