360 Secure SkillHub
360 安全技能中心

YouTube Archiver

把 YouTube 播放列表（如 Liked、Watch Later）导入为带元数据和转录的 Markdown 笔记，并可选调用 LLM 生成摘要与标签；支持按 video_id 去重、锁文件防并发，以及用 cron 做定时同步。

benmillerat
benmillerat
热度 60.06
最近更新 2026-05-11 17:27:20
分类：知识依赖：代码执行 / 读写文件 / 第三方 API / 内部调用大模型应用场景：笔记整理 / 摘要总结 / 自动化脚本 / 工作流

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add youtube-archiver
Metadata
name
youtube-archiver
description
Archive YouTube playlists into markdown notes with metadata, transcripts, AI summaries, and tags. Use when a user asks to import/sync YouTube playlists, archive Watch Later or Liked videos, enrich YouTube notes, batch process video notes, or automate recurring YouTube-to-markdown sync jobs with cron.

YouTube Archiver

Use this skill to import YouTube playlists into markdown files and optionally enrich notes with transcript, summary, and tagging.

Requirements

  • Python 3.7+
  • yt-dlp (pip install yt-dlp or brew install yt-dlp)
  • A browser signed into YouTube (for private playlists like Liked/Watch Later)
  • macOS: terminal needs Full Disk Access to read browser cookies
  • Windows: browser cookie extraction can be flaky; cookies_file export is the safer path
  • Linux: works on desktop installs; headless servers need cookies_file

First-run setup flow (interactive)

If no config exists at <output>/.config.json, ask these questions before running scripts.

Required questions

  1. Where should archived notes be stored?
    • Default: ./YouTube-Archive
  2. Which playlists should be archived?
    • Accept playlist IDs or URLs
    • Default: LL (Liked Videos), WL (Watch Later)
  3. Which browser is signed into YouTube for cookie auth?
    • Default: chrome

Optional enrichment questions

Ask only if the user wants summaries/tags.

  1. Generate AI summaries? (yes/no)
  2. Summary provider? (openai, gemini, anthropic, openrouter, ollama, none)
  3. Summary model name?
  4. API key env var name?
  5. Enable auto-tagging? (yes/no)
  6. Tagging provider/model/env var?
  7. Keep default tags or define custom vocabulary?

First-run execution sequence

  1. Run init:
    • python3 <skill>/scripts/yt-import.py --output <output-dir> --init
  2. Edit <output-dir>/.config.json from the user’s answers.
  3. Verify auth with dry run:
    • python3 <skill>/scripts/yt-import.py --output <output-dir> --dry-run
  4. Run real import.
  5. Run enrichment (optional):
    • python3 <skill>/scripts/yt-enrich.py --output <output-dir> --limit 10

One-shot quick start

Use this for immediate manual sync:

python3 <skill>/scripts/yt-import.py --output <output-dir>
python3 <skill>/scripts/yt-enrich.py --output <output-dir> --limit 10

Useful import flags:

  • --dry-run
  • --playlist <ID> (repeatable)
  • --no-summary
  • --no-tags
  • --cookies <path/to/cookies.txt>
  • --browser <name>

Useful enrich flags:

  • --dry-run
  • --limit <N>
  • --strict-config

Idempotency and safety behavior

  • Import skips already archived videos by video_id.
  • Filenames include video ID: Title [video_id].md.
  • Enrichment skips notes where frontmatter has enriched: true.
  • Lockfile prevents concurrent runs: <output-dir>/.yt-archiver.lock.

Automation with cron (single-agent default)

Offer cron only after one successful manual run.

Example schedule (daily 11:00):

  1. Import new videos
  2. Enrich a bounded batch

Example task text:

  • Run yt-import.py for <output-dir>, then run yt-enrich.py --limit 10 for the same output.

Keep it single-agent by default. Do not assume multi-agent routing.

Troubleshooting and provider details

Read these references when needed:

  • Provider setup, model suggestions, cost: references/providers.md
  • Common failures and fixes: references/troubleshooting.md
  • Default summary prompt template: references/default-summary-prompt.md