jarodise How To Read a Book — NotebookLM Reading Companion
Overview
Transform EPUB-formatted ebooks into chapter-level Markdown sources for NotebookLM, paired with a custom "How To Read a Book" analytical reading persona. This skill orchestrates the complete workflow: parsing EPUB structure, splitting by chapters, creating a NotebookLM notebook, injecting the reading companion persona, and uploading all sources.
When to Use
Use this skill when the user wants to:
- Upload an EPUB book to NotebookLM for deep reading
- Create a structured reading companion for non-fiction books
- Split an EPUB into chapters and upload them as separate sources
- Apply analytical reading methodology to a book in NotebookLM
- Transform ebooks into NLP-friendly chapter format for AI analysis
Prerequisites
Before running this skill, ensure:
- The user has
notebooklm-pyinstalled:pip install notebooklm-py playwright - Playwright installed:
playwright install chromium - User is authenticated:
notebooklm login - User has an EPUB file locally available
How to Use
Step 1: Verify Environment
Check if dependencies are installed and user is authenticated:
which notebooklm
notebooklm list # Should succeed if authenticated
If not installed, guide user through setup or run ./install.sh if it exists.
Step 2: Get EPUB Path
Ask the user for the path to their EPUB file, or confirm the path if they provided it.
User prompts that indicate EPUB path:
- "I have a book at /path/to/book.epub"
- "Process this EPUB: ~/Downloads/mybook.epub"
- "Upload this book to NotebookLM" (ask for path)
Step 3: Run the Skill
Execute the main script with the EPUB path:
python3 /Users/jarodise/.claude/skills/how-to-read-a-book/scripts/run.py <path-to-epub>
Step 4: Report Results
The script will output:
- Progress indicators for each phase (📖 Parsing → ✍️ Converting → 📤 Uploading → ✅ Complete)
- JSON result with
notebook_id,notebook_url,book_title,chapter_count,sources_uploaded - Human-readable summary with NotebookLM link
Step 5: Guide the User
Once complete, tell the user:
- How many chapters were uploaded
- The NotebookLM notebook URL
- That the reading companion persona is now active
- Suggest starting questions like "What is the main problem this book is trying to solve?" or "Summarize Chapter 1's key argument"
What the Skill Does
- Parse EPUB: Extract book metadata and chapter structure using TOC-first with heading-based fallback
- Split into Chapters: Convert each chapter to Markdown with frontmatter (number, title, source book)
- Create Notebook: Create new notebook titled
<Book Title> — Reading Companion - Configure Persona: Inject "How To Read a Book" analytical reading methodology as system prompt
- Upload Sources: Upload all chapter Markdown files as individual sources
- Cleanup: Remove temporary files after successful upload
Error Handling
| Error | Response |
|---|---|
| EPUB file not found | "I couldn't find that EPUB file. Please check the path: <path>" |
| Invalid/corrupt EPUB | "This EPUB appears to be damaged or in an unsupported format." |
| NotebookLM not installed | "Please install dependencies first: pip install notebooklm-py playwright && playwright install chromium" |
| Not authenticated | "Please authenticate with NotebookLM first: notebooklm login" |
| Upload fails | Report partial success, preserve temp files for retry |
Expected Output Format
The script outputs JSON after ---JSON_OUTPUT--- delimiter:
{
"notebook_id": "uuid-here",
"notebook_url": "https://notebooklm.google.com/notebook/uuid-here",
"book_title": "Book Title Here",
"chapter_count": 12,
"sources_uploaded": 12
}
Limitations
- EPUB format only (not PDF, MOBI, or AZW)
- Local files only (no URL downloading)
- Single universal reading companion persona (no book-type variations)
- Post-upload conversation happens in NotebookLM interface, not via this skill
Reference
- Pattern based on:
/Users/jarodise/Documents/GitHub/CNinfo2Notebookllm - EPUB parsing:
ebooklibPython library - NotebookLM CLI:
notebooklm-py