AI Summary Skill for Claude Code

A Claude Code skill that generates illustrated, structured summary reports from PDFs, web pages, YouTube videos, and audio/video files — with Mermaid diagrams, anti-hallucination verification, and quality self-assessment.

Features

• Multi-source input: PDF, DOCX, TXT, web URLs, YouTube, audio/video
• AI-powered summarization: Claude generates structured summaries with fact verification
• Illustrated output: Every chapter includes a Mermaid diagram (pie charts, flowcharts, graphs)
• Anti-hallucination: 4-layer defense — fact pre-extraction, outline validation, per-chapter source checking, quality self-assessment
• Zero server deployment: No backend needed — Claude does the AI work directly
• Progressive capability: Works with zero configuration, unlocks more features as you add API keys

Quick Start

1. Install

Option A — Project-level install (recommended):

# Clone the repo
git clone https://github.com/molly554/Multi_sources_summarizition_lynote.ai.git

# Copy the Skill into your project's Claude Code skills directory
# Note: .claude/skills/ is relative to your Claude Code project root
cp -r Multi_sources_summarizition_lynote.ai/.claude/skills/summarize YOUR_PROJECT/.claude/skills/summarize

How to find your project root? Open Claude Code and check the path shown in the bottom status bar. For example, if it shows C:\Users\you\myproject, copy to C:\Users\you\myproject\.claude\skills\summarize\

Option B — Global install (available in all projects):

# macOS / Linux
cp -r Multi_sources_summarizition_lynote.ai/.claude/skills/summarize ~/.claude/skills/summarize

# Windows (Git Bash)
cp -r Multi_sources_summarizition_lynote.ai/.claude/skills/summarize "$USERPROFILE/.claude/skills/summarize"

Option C — Verify installation:

# Start Claude Code, type /summarize
# If autocomplete shows the skill, installation is successful

2. Use

# Start Claude Code in your project
claude

# Summarize a PDF
/summarize report.pdf

# Summarize a web page
/summarize https://example.com/article

# Summarize multiple sources
/summarize paper.pdf https://example.com https://youtube.com/watch?v=xxx

Capability Levels

Level 1: Zero Config (no API keys needed)

Input	How it works
PDF (≤5 pages)	Claude reads directly — text + visual understanding of charts/tables
DOCX / TXT	Claude reads directly
Web URLs	WebFetch extracts article content

What you get: Structured Markdown summary + Mermaid diagrams + quality score

Level 2: Large Documents

Requires: AZURE_DI_ENDPOINT + AZURE_DI_KEY

Unlocks: PDF files with more than 5 pages, processed via Azure Document Intelligence for high-speed, high-accuracy extraction with table/layout recognition.

Level 3: YouTube Videos

Requires: SUPADATA_API_KEY

Unlocks: YouTube video transcript extraction and summarization.

Level 4: Audio & Video Files

Requires: DEEPGRAM_API_KEY

Unlocks: Local MP3, MP4, WAV file transcription via Deepgram Nova-3.

Configuration

Copy the example env file and fill in the keys you need:

cp .env.example .env

Then set the environment variables before starting Claude Code:

# Option A: Export in your shell
export AZURE_DI_ENDPOINT="https://your-instance.cognitiveservices.azure.com"
export AZURE_DI_KEY="your_key_here"
export SUPADATA_API_KEY="your_key_here"
export DEEPGRAM_API_KEY="your_key_here"

# Option B: Add to your shell profile (~/.bashrc, ~/.zshrc)
# Option C: Use direnv with .envrc

Getting API Keys

Service	Free Tier	Sign Up
Azure Document Intelligence	500 pages/month free	Azure Portal → Create "Document Intelligence" resource
Supadata (YouTube transcripts)	50 requests/month free	supadata.ai
Deepgram (Audio transcription)	$200 free credit	deepgram.com

Output

Every summary generates two files:

File	Purpose	How to view
./output/summary-{timestamp}.md	Source Markdown	VSCode, GitHub, Typora
./output/summary-{timestamp}.html	Visual report	Double-click → opens in browser

The HTML file is a standalone page with embedded Mermaid JS rendering — no server or plugins needed. Just double-click to see the fully illustrated report.

Both files contain:

• Title and source attribution
• Executive summary
• 3-6 structured chapters, each with:
  • Sub-sections with key data points
  • Mermaid diagram (pie chart, flowchart, or graph)
• Conclusions and recommendations
• Quality self-assessment score (D1-D4)

Example Output Structure

# Core Insights: AI Market Trends 2026

**Source documents:**
- ai-trends-report.pdf
- https://techcrunch.com/2026/03/ai-funding

**Summary**
The AI market is projected to reach $500B by 2027...

## 1. Market Overview

### 1.1 Growth Trajectory
Key findings from the report show **47% YoY growth**...

\`\`\`mermaid
pie title AI Market Share 2026
    "Enterprise AI" : 45
    "Consumer AI" : 30
    "AI Infrastructure" : 25
\`\`\`

### 1.2 Regional Distribution
...

## 2. Investment Landscape
...

## Conclusions and Recommendations
1. ...
2. ...
3. ...

Security

• Zero hardcoded secrets: No API keys or server URLs in the skill code
• Direct API calls: Your keys go straight to Azure/Supadata/Deepgram — no intermediary servers
• Local processing: Claude does AI work locally in your session — no data leaves your machine except to the APIs you configure
• No backend required: Nothing to deploy, no server to maintain

How It Works

/summarize input1 input2 ...
         │
         ▼
  Phase 0: Parse inputs → classify as PDF/URL/YouTube/Audio
         │
         ▼
  Phase 1: Extract content
         │  PDF ≤5pg → Claude Read (free)
         │  PDF >5pg → Azure DI (or fallback to Claude Read)
         │  URL → WebFetch
         │  YouTube → Supadata API
         │  Audio → Deepgram API
         │
         ▼
  Phase 2: AI Summary (Claude does this directly)
         │  Step A: Extract facts from source text
         │  Step B: Generate structured outline
         │  Step C: Validate outline against facts
         │  Step D: Write chapters + Mermaid diagrams
         │
         ▼
  Phase 3: Output
         │  Assemble Markdown
         │  Save to ./output/
         │  Quality self-assessment
         │  Display preview
         ▼
  Done!

Powered by LyNote.ai

Like this Skill? Try the full product — no setup required.

LyNote.ai — All-in-One AI Learning Assistant that summarizes YouTube videos, PDFs, articles, podcasts, and more. Instantly create illustrated notes and boost your learning efficiency by 10x.

What's coming next:

Feature	Description
AI Research Assistant	Summarize videos & articles, transcribe audio, ask questions to your documents
AI Content Studio	Generate outlines, drafts, presentations — from idea to finished product
AI Study Partner	Flashcards, quizzes, and exam prep from any learning material
Knowledge Hub	Organize notes, collaborate in real-time, share securely

Supported content: YouTube | Podcast | PDF | Audio | Video | Web Articles | Books

Add to Chrome - Free →

Join thousands of students, professionals, and lifelong learners who are saving time and unlocking knowledge with LyNote.

License

MIT