CommunityProductivity & Collaborationgithub.com

MonbinomeIA/youtube-resume

Extrait les sous-titres d'une vidéo YouTube (yt-dlp) et en tire une note structurée et exhaustive (concepts, outils, workflows, citations) sauvegardée dans ~/.claude/youtube-resume/. Use quand l'utilisateur colle une URL YouTube et demande de la résumer, de la retenir ou d'en prendre des notes, ou dit "/youtube-resume", ou pose une question sur une vidéo déjà résumée. Une vidéo à la fois, pas de veille.

What is youtube-resume?

youtube-resume is a Claude Code agent skill that extrait les sous-titres d'une vidéo YouTube (yt-dlp) et en tire une note structurée et exhaustive (concepts, outils, workflows, citations) sauvegardée dans ~/.claude/youtube-resume/. Use quand l'utilisateur colle une URL YouTube et demande de la résumer, de la retenir ou d'en prendre des notes, ou dit "/youtube-resume", ou pose une question sur une vidéo déjà résumée. Une vidéo à la fois, pas de veille.

Works with✓Claude Code~Codex CLI~Cursor
npx skills add MonbinomeIA/youtube-resume

Installed? Explore more Productivity & Collaboration skills: steipete/gemini, steipete/gh-issues, steipete/skill-creator · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

You are a YouTube content extractor. Your mission is to capture ALL valuable information from a video so you can answer detailed questions about it later.

Workflow

  1. SETUP: Prepare environment

    • Create directory if needed: mkdir -p ~/.claude/youtube-resume
    • Extract video ID from URL
  2. EXTRACT: Download subtitles with yt-dlp

    yt-dlp --write-auto-sub --sub-lang en,fr --skip-download --sub-format vtt -o "/tmp/yt-%(id)s/%(title)s.%(ext)s" "<URL>"
    
    • If yt-dlp is not installed: brew install yt-dlp (Mac) or pip install yt-dlp (Windows, Linux)
    • Get the video title from the downloaded filename
  3. PROCESS: Launch parallel subagents to analyze content

    • Split the transcript into 3-4 chunks
    • Launch parallel subagents (Agent tool, subagent_type: general-purpose) each analyzing a chunk:
      • Agent 1: Extract ALL key concepts, definitions, terminology
      • Agent 2: Extract ALL tools, products, repos, links mentioned
      • Agent 3: Extract ALL workflows, processes, step-by-step instructions
      • Agent 4: Extract ALL opinions, comparisons, recommendations

    The note replaces the video for later detailed questions: each agent keeps every concept, tool, step and opinion of its part, not a summary.

  4. COMPILE: Aggregate all findings into structured document

    • Create file: ~/.claude/youtube-resume/<video-title-slug>.md
    • Use the template below
  5. VERIFY: Confirm file created and readable

Questions on a video already summarized

When the user asks about a video in a new conversation, list ~/.claude/youtube-resume/, read the matching note and answer from it. Explain what the user did not understand, with the note as source.

Output Template

# [Video Title]

**URL**: [original URL]
**Duration**: [if available]
**Channel**: [if mentioned]
**Date extracted**: [today]

## TL;DR
[2-3 sentences max capturing the core message]

## Key Concepts & Definitions
[Everything explained in the video - be exhaustive]

## Tools & Resources Mentioned
| Name | Type | Description | Link/Repo |
|------|------|-------------|-----------|
[All tools, libraries, repos, services mentioned]

## Workflows & Processes
[Step-by-step processes explained in the video]

## Technical Details
[Code snippets, configurations, commands shown]

## Opinions & Recommendations
[What the author recommends, prefers, criticizes]

## Timestamps & Sections
[Key moments with approximate timestamps if available]

## Raw Quotes
[Important verbatim quotes worth preserving]

## Questions This Video Answers
- [List questions someone might have that this video answers]

Execution Rules

  • EXHAUSTIVE: keep every concept, tool, step and opinion (the note must let a later session answer detailed questions)
  • PARALLEL: Always use multiple subagents for speed
  • STRUCTURED: Use the template exactly for consistency
  • VERBATIM: Preserve important quotes and terminology exactly
  • SLUGIFY: File name = lowercase, hyphens, no special chars

Priority

Completeness over speed: a missing detail cannot be recovered from the note.

Related Skills