| name | notebooklm-splitter |
| description | Skill to audit a large NotebookLM notebook and split its sources into multiple specialized notebooks based on identified themes. |
| license | Apache-2.0 |
| metadata | {"version":"1.0","capabilities":["auditing","content-extraction","notebook-creation","specialization"]} |
NotebookLM Splitter Skill
This skill allows an agent to decompose a heterogeneous "mega-notebook" into a set of specialized, high-signal notebooks.
Workflow
1. Source Inventory & Thematic Analysis
- Use
notebook_get or browser to list all sources.
- Group sources by topic, format, or target audience.
- Define a "Taxonomy of Knowledge" for the notebook.
2. Specialized Blueprinting
- Propose a set of new notebooks. Each must have:
- A clear, specific Title.
- A defined Goal.
- A list of sources to be moved/copied.
3. Execution (The Split)
- Create each specialized notebook using
notebook_create.
- For each source in the group:
- Extract text content via
source_get_content.
- Inject into the new notebook via
notebook_add_text.
- Optional: Reference the original source URL if available via
notebook_add_url.
4. Persona Tailoring
- For each new notebook, configure a custom persona (
configure_chat) that focuses strictly on the specialized scope.
5. Final Audit
- Verfiy that the original notebook is now a "Master Index" or can be archived.
- Verify that each new notebook provides better retrieval quality than the original.
Best Practices
- Fragment Control: Don't split too thin. Aim for 10-30 high-quality sources per specialized notebook.
- Context Injection: When adding text sources, include a header indicating the original source name and date.
- Goal Alignment: The persona must be the "Subject Matter Expert" for that specific split.
