Claude Code Skill

/live-workshop Skill

Turn any lecture into a structured, searchable knowledge base — in real time.

Four-Phase Pipeline

From raw audio to organized, publishable knowledge

⚙
Phase 1

INIT

  • Creates JupyterBook project structure
  • Auto-fetches agenda from workshop URL
  • Git repo with proper .gitignore
  • Modular: theory / tools / exercises per topic
♫
Phase 2

RECORD

  • Live audio capture via sox (configurable chunks)
  • Real-time Whisper transcription (tiny→turbo)
  • Multi-language support (default: Polish)
  • Keyword extraction with custom categories
📂
Phase 3

ORGANIZE

  • Auto-classifies notes into correct module
  • Updates daily journal with structured entries
  • Cross-references via {doc} directives
  • Git commits after every meaningful update
🧠
Phase 4

ANALYZE

  • Iterative transcript analysis on demand
  • Decodes Whisper transcription errors
  • Extracts pedagogical insights from speaker
  • Builds error tables for quality assessment

Key Features

Everything you need for real-time workshop documentation

📖

JupyterBook Output

Generates a publishable HTML book with table of contents, cross-references, and structured chapters. Ready to host on GitHub Pages.

█

Live Terminal Prompter

Real-time colored keyword tags stream to your terminal as audio is processed. Instant visual feedback on what the speaker is covering.

🔭

Selective Keyword Detection

Customizable keyword categories per workshop topic. Filter signal from noise with domain-specific vocabulary lists.

🇧🇱

~70-80% Non-English Accuracy

Optimized for real-world non-English transcription. Handles mixed-language lectures, technical jargon, and accented speech.

🔁

Automatic Deduplication

Detects and skips duplicate transcript analysis. No wasted computation when re-processing previously analyzed fragments.

🔒

Works Fully Offline

Local Whisper models, no cloud API needed. Your audio and transcripts never leave your machine. Complete data sovereignty.

Quick Start
/live-workshop "Cybersecurity 101" --url https://workshop.example.com --lang en

▾ How to Install This Skill

  1. Install Claude Code if you don't have it yet: npm install -g @anthropic-ai/claude-code
  2. Clone the skill into your Claude skills directory: git clone https://github.com/nasqret/live-workshop-skill.git ~/.claude/skills/live-workshop
    Or manually copy the SKILL.md and templates/ folder into ~/.claude/skills/live-workshop/
  3. Install audio dependencies: brew install sox ffmpeg && pip3 install openai-whisper jupyter-book
    On Linux: apt install sox ffmpeg
  4. Navigate to any project directory and invoke: /live-workshop "Your Workshop Title" --lang en --modules "Topic1,Topic2"