youtube-archiver_skill

Overview

This skill archives YouTube playlists into local markdown notes with metadata, transcripts, AI summaries, and tags. It imports playlists (including Watch Later and Liked), keeps notes idempotent, and can enrich existing notes with transcripts, summaries, and automatic tagging. Use it to back up or batch-process YouTube content into a searchable, human-readable note collection.

How this skill works

The importer uses yt-dlp and your browser cookies (or a cookies file) to fetch playlist and video metadata, downloading transcripts when available and creating one markdown file per video named with the video ID. An optional enrichment step calls a configured AI provider to generate summaries and tags, and writes frontmatter flags so files are skipped if already enriched. A lockfile and idempotent checks prevent duplicates and concurrent runs.

When to use it

• Import or sync one or more YouTube playlists into markdown
• Archive Watch Later, Liked, or private playlists for backup
• Enrich existing video notes with transcripts, AI summaries, and tags
• Batch-process many videos or set up recurring cron sync jobs
• Quick one-shot exports for migration or offline access

Best practices

• Run the interactive init once to create a stable .config.json before scheduling automation
• Use a cookies file for reliable auth on servers; desktop browsers work for local runs
• Start with --dry-run to verify config and playlist access before full imports
• Limit enrichment batches (e.g., --limit 10) to control API costs and rate limits
• Keep API keys in environment variables and choose provider/model appropriate to cost and latency

Example use cases

• Daily cron job: import new playlist videos then enrich a capped batch for summaries and tags
• Migrate a creator’s public playlist into a personal knowledge base with transcripts and metadata
• Bulk-archive Watch Later before clearing it from YouTube, preserving captions and timestamps
• Periodic enrichment pass to add improved AI summaries or switch providers/models across existing notes