notebooklm

The NotebookLM Research Assistant skill acts as a specialized bridge between Claude Code and Google NotebookLM, enabling users to perform source-grounded, citation-backed research directly within their development environment. By utilizing an automated headless browser orchestration layer, it retrieves context specifically from your uploaded documents to minimize hallucinations and ensure responses are strictly confined to the provided source material. This skill is ideal for developers, researchers, and technical writers who maintain extensive documentation in NotebookLM and require a reliable way to query that knowledge base without switching contexts between their IDE and browser.

• Perform source-grounded queries against specific NotebookLM notebooks using Gemini as the underlying engine.

• Manage your notebook library including listing, searching by topic, activating, and removing notebooks via CLI commands.

• Automated Smart Discovery: If a notebook lacks metadata, the skill can programmatically analyze its contents to suggest appropriate titles, descriptions, and tag structures.

• Integrated authentication management with persistent sessions, providing a secure way to handle Google login flows via a visible browser window.

• Follow-up mechanism that enforces rigorous verification: after each answer, the skill identifies gaps and prompts for additional context to ensure comprehensive results.

• The skill requires a local Python environment; always use the 'python scripts/run.py' wrapper to execute commands, as it handles virtual environment activation and dependency management automatically.

• When adding a notebook, provide the URL, a descriptive name, a clear description, and relevant topics. Use the 'Smart Add' command if you are unsure of the categorization.

• Always check authentication status using 'auth_manager.py status' before initiating queries to ensure the session is active.

• The tool is designed for structured interaction: every query sequence concludes with a verification check to ensure all user information needs are fully satisfied through repeated, grounded questioning where necessary.