NotebookLM — Browser Automation

This skill automates Google's NotebookLM through browser automation, enabling actions like querying notebooks, adding sources, generating Studio outputs (Audio/Video Overviews, Mind Maps, Reports, etc.), and creating new notebooks. It handles UI discovery, async generation, and provides structured intake questions to route actions correctly.

✨ What it does

  • Guides the user through a dependency-ordered intake (Grill-Me) to determine the action, notebook, and parameters.
  • Performs browser automation with screenshot-first discipline and semantic element finders.
  • Handles adding sources via multiple sub-flows (URL, text, file, Google Doc, synthesized content).
  • Generates Studio outputs with mandatory custom prompts for quality, discovering available output types from the live UI.
  • Implements fire-and-notify for long-running Studio generations (e.g., Audio Overview) to avoid session timeouts.
  • Provides error handling for missing browser automation and login walls.

🎯 When to use it

  • When the user wants to ask a question of an existing NotebookLM notebook.
  • When the user wants to add a source (URL, text, file, Google Doc, or synthesized content) to a notebook.
  • When the user wants to generate a Studio output such as Audio Overview, Mind Map, or Study Guide from a notebook.
  • When the user wants to create a new notebook in NotebookLM.

🚀 How to use

Trigger the skill by asking to do something in NotebookLM, e.g., 'open NotebookLM', 'ask my notebook about X', 'add this URL to NotebookLM', or 'generate a Video Overview from my notebook'. The skill will ask up to four forcing questions to route the action. It requires a browser automation environment (e.g., Claude Code CLI with computer-use). Example prompts:

Ask my notebook about the key findings in the Q3 report.
Add this article URL to my research notebook.
Generate an Audio Overview from my notes for a non-technical audience.

📄 Output: The skill produces the requested action result: a chat answer, added source confirmation, a generated Studio output (hosted in NotebookLM), or a new notebook, with a clean summary report.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/research/notebooklm/skills/notebooklm .claude/skills/notebooklm

Skill source: research/notebooklm/skills/notebooklm/SKILL.md

⚠️ Good to know

Requires a browser automation environment; fails gracefully with a clear message if unavailable, and never attempts to handle login automatically.

❓ FAQ

What happens if I ask to 'open NotebookLM' without specifying an action?

The skill will refuse to start and re-ask the first intake question to determine the desired action.

Why does the skill always open the customization menu for Studio outputs?

Because default prompts produce mediocre output; a detailed custom prompt is mandatory for quality.

Does the skill wait for Studio generations like Audio Overview to complete?

No, for long-running generations it uses a fire-and-notify pattern: it confirms generation started and tells the user they will be notified when ready.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.