NotebookLM Lesson Planner

Create comprehensive lesson plans for Google NotebookLM with validated sources and customized audio overview prompts.

What This Skill Does

This skill helps you create NotebookLM notebooks by:

•Finding high-quality sources - Identifying 10-20 online resources that are publicly accessible, LLM-readable, and substantive
•Validating sources - Ensuring all URLs are active, accessible, and not behind paywalls or crawler blocks
•Generating tailored prompts - Creating specific prompts for each learning area that guide NotebookLM's audio overview generation
•Structuring content - Organizing lessons to avoid redundancy and ensure comprehensive coverage

When to Use This Skill

Use this skill when the user wants to:

•Create a NotebookLM notebook on a specific topic
•Generate a learning curriculum with multiple focused audio overviews
•Build educational content with curated sources
•Process multiple topics in batch (e.g., 100+ lesson plans)

Workflow Overview

The process varies based on whether you're creating a single lesson or processing a batch:

Single Lesson Mode:

•Gather requirements from user
•Find and validate sources
•Generate prompts for each learning area
•Provide complete lesson plan

Batch Mode:

•Process JSON specification file
•For each lesson: find sources, validate, generate prompts
•Output complete lesson plans for all topics

Single Lesson Mode

Step 1: Gather Requirements

Ask the user for:

•Main topic - The overall subject (e.g., "Firewall Configuration")
•Learning areas - Specific subtopics to cover (e.g., "ACLs", "NAT", "VPN", "Logging")
•
Listener context - Information about the learner:
- •What they already know
- •Expertise level (beginner, intermediate, advanced)
- •What they want to learn
- •Why they're learning this (use case, goal)
•Target source count - How many sources to find (recommend 15, range 10-20)

Step 2: Find Sources

Use web_search to find sources that:

•Cover the main topic and learning areas
•Are publicly accessible (no paywall, no login)
•Are high-quality and substantive
•Provide appropriate depth for listener's level

Search strategy:

•Start with comprehensive sources covering multiple learning areas
•Find specialized sources for each specific learning area
•Ensure diversity in source types (tutorials, documentation, analysis)
•Aim for 10-20 total sources

See references/source_finding.md for detailed strategies on:

•Source quality criteria
•Search techniques
•Source type evaluation
•Coverage matrix planning

Step 3: Validate Sources

After finding sources, validate them using the validation script:

bash

scripts/validate_sources.py url1 url2 url3 ... --verbose

The script checks:

•URL is active (200 status)
•No paywall detected
•Content is substantive (>500 chars)
•No obvious crawler blocking

If sources fail validation:

•Remove invalid sources
•Find replacements using web_search
•Re-validate until all sources pass

Step 4: Generate Prompts

For each learning area, create a tailored prompt for NotebookLM's audio overview generator.

Read prompt templates first:

bash

view references/prompt_templates.md

Each prompt must include:

•Topic/Learning Area - What specific area this overview covers
•Listener Profile - Who the listener is, what they know, their expertise level
•Coverage Scope - What this overview should cover in detail
•Previous Coverage - What was covered in the last episode (for continuity)
•Upcoming Topics - What will be covered in future episodes (to avoid going into detail)
•Exclusions - Critical! What NOT to cover in detail (handled by other overviews)
•Approach and Goals - Tone, depth, teaching style, and learning objectives
•Length Suggestion - Duration target: [~15 min] for standard, [~22 min] for longer episodes
•Additional Context - Any other relevant information

Exclusion and continuity strategy:

•Overview 1: No previous coverage, upcoming: areas 2, 3, 4
•Overview 2: Previous: area 1, upcoming: areas 3, 4
•Overview 3: Previous: areas 1-2, upcoming: area 4
•Overview 4: Previous: areas 1-3, no upcoming topics

This ensures each overview is focused, non-redundant, and builds on prior knowledge.

Case Study Requirement: One audio overview in every lesson MUST be a real-life case study focusing on real companies and their experiences. This overview should analyze actual implementation stories, challenges faced, and lessons learned from real organizations.

Step 5: Generate Quiz Prompts

For each lesson, create 4 quiz prompts based on the audio overviews. Each quiz prompt should:

•Reference the audio overview - Which overview this quiz is based on
•Specify length - short, default, or long
•Specify difficulty - medium or hard
•Guide quiz content - What topics/concepts to focus on

Quiz distribution:

•Quiz 1: Based on Overview 1 (difficulty: medium, length: default)
•Quiz 2: Based on Overview 2 (difficulty: medium, length: short)
•Quiz 3: Based on Overview 3 (difficulty: hard, length: default)
•Quiz 4: Based on Overview 4 or case study (difficulty: hard, length: long)

Step 6: Deliver Lesson Plan

Provide the user with:

•Complete source list - All validated URLs with brief descriptions
•Coverage summary - Which sources cover which learning areas
•Audio Overview Prompts - One prompt per learning area, numbered and ready to use (including length suggestions)
•Quiz Prompts - Four quiz prompts with length and difficulty specified
•
Instructions - How to use the lesson plan with NotebookLM:
- •Add all sources to NotebookLM notebook
- •For each audio overview, copy the corresponding prompt
- •Generate audio overviews in sequence
- •Generate quizzes using the quiz prompts

Output format:

markdown

# NotebookLM Lesson Plan: [Topic]

## Sources (15)

1. [Source 1 Title] - [URL]
   Coverage: [Learning areas covered]

2. [Source 2 Title] - [URL]
   Coverage: [Learning areas covered]

...

## Audio Overview Prompts

### Overview 1: [Learning Area 1] [~15 min]

[Full prompt text including: coverage, previous coverage, upcoming topics, approach and goals]

### Overview 2: [Learning Area 2] [~15 min]

[Full prompt text]

### Overview 3: [Case Study] [~22 min]

[Full prompt text - real-life case study of real companies]

...

## Quiz Prompts

### Quiz 1: [Topic] (Based on Overview 1)
**Length**: default | **Difficulty**: medium

[Full quiz prompt text]

### Quiz 2: [Topic] (Based on Overview 2)
**Length**: short | **Difficulty**: medium

[Full quiz prompt text]

...

## Usage Instructions

1. Create a new notebook in Google NotebookLM
2. Add all 15 sources to the notebook
3. For each audio overview:
   - Click "Generate Audio Overview"
   - Copy and paste the corresponding prompt above
   - Generate the overview
4. For each quiz:
   - Click "Generate Quiz"
   - Copy and paste the corresponding quiz prompt
   - Generate the quiz
5. Listen to overviews in sequence (1→2→3→...)

Batch Mode

For processing multiple topics from a JSON file.

Step 1: Understand Batch Specification

The user provides a JSON file with this structure:

json

{
  "lessons": [
    {
      "topic": "Sandwich Making",
      "learning_areas": ["bread", "proteins", "cheeses", "vegetables"],
      "listener_context": {
        "knows": ["basic cooking"],
        "wants_to_learn": ["gourmet techniques"],
        "skill_level": "intermediate"
      },
      "target_sources": 15
    },
    {
      "topic": "Firewall Configuration",
      "learning_areas": ["ACLs", "NAT", "VPN"],
      ...
    }
  ]
}

Step 2: Process Each Lesson

For each lesson in the batch:

•Follow the Single Lesson Mode workflow
•Find sources (may need multiple web_search calls)
•Validate sources
•Generate prompts
•Save lesson plan to individual JSON file

Important for batch processing:

•Process lessons sequentially (don't try to parallelize)
•Be efficient with web_search (combine queries when possible)
•Validate sources in batches using the script
•Save progress incrementally (one lesson at a time)

Step 3: Output Structure

Create a directory structure:

code

output/
├── lesson_001_sandwich_making.json
├── lesson_002_firewall_configuration.json
├── lesson_003_database_optimization.json
...
└── batch_summary.md

Each JSON file contains:

json

{
  "topic": "Sandwich Making",
  "learning_areas": ["bread", "proteins", "cheeses", "vegetables"],
  "listener_context": {...},
  "sources": [
    {
      "url": "https://example.com/sandwich-guide",
      "title": "Complete Sandwich Making Guide",
      "coverage": ["bread", "proteins", "cheeses", "vegetables"]
    },
    ...
  ],
  "prompts": [
    {
      "area": "bread",
      "prompt": "Topic: Bread Selection for Sandwiches\n\n..."
    },
    ...
  ]
}

The batch_summary.md provides an overview of all lessons created.

Quality Guidelines

Source Quality

•Minimum 10 sources, maximum 20
•All sources must pass validation
•Each learning area covered by at least 3 sources
•Mix of comprehensive and specialized sources
•Appropriate depth for listener's expertise level

Prompt Quality

•Each prompt 200-350 words (optimal with new requirements)
•Clear previous coverage and upcoming topics sections
•Clear exclusions to avoid overlap
•Listener context consistent across all prompts
•Specific coverage guidance
•Appropriate tone and depth for audience
•Length suggestion included ([~15 min] or [~22 min])
•Approach and goals clearly stated
•At least one case study overview per lesson

Quiz Quality

•4 quizzes per lesson minimum
•Each quiz mapped to a specific audio overview
•Mix of difficulty levels (medium and hard)
•Mix of lengths (short, default, long)
•Clear guidance on topics to test
•Questions should reinforce key concepts from the audio overviews

Coverage Balance

•No learning area should dominate (unless intentional)
•No learning area should be under-represented
•Sources should provide diverse perspectives
•Both breadth and depth appropriate for topic scope

Tips and Best Practices

Finding sources:

•Start broad, then get specific
•Use multiple search queries per learning area
•Look for official documentation first
•Check publication date for time-sensitive topics
•Verify accessibility before adding to list

Writing prompts:

•Be explicit about exclusions
•Reference other overviews by number
•Provide specific examples of what to cover
•Match technical depth to listener's level
•Include practical context (why they're learning this)

Validating sources:

•Always run validate_sources.py before finalizing
•Manually spot-check 3-5 sources
•If >20% fail validation, reconsider search strategy
•Replace failed sources before generating prompts

Batch processing:

•Process 10-20 lessons at a time
•Check one complete lesson before processing all
•Save progress frequently
•Use consistent listener_context format across batch

Common Issues

Sources failing validation:

•Check for paywall indicators in URL or title
•Try alternative sources on the same topic
•Ensure using HTTPS not HTTP
•Verify site isn't down temporarily

Overlapping prompts:

•Review exclusion sections
•Ensure each area has clear boundaries
•Consider if learning areas need restructuring
•Make exclusions more specific

Coverage gaps:

•Add more specialized sources
•Use more specific search queries
•Look for sources that bridge multiple areas
•Consider if learning areas are too granular

Batch processing slow:

•Combine search queries where possible
•Validate sources in batches
•Process during off-peak hours
•Consider breaking into smaller batches

Files and Scripts

Scripts:

•validate_sources.py - Validate URLs for NotebookLM compatibility

References:

•prompt_templates.md - Templates and examples for creating prompts
•source_finding.md - Strategies for finding high-quality sources

Usage examples in references files.