openclaw-notebooklm-auditor

name	openclaw-notebooklm-auditor
description	Specialized skill to audit, optimize, and maintain the NotebookLM workspace for "OpenClaw: The Universal Personal AI Assistant Platform". Use this skill when you need to: (1) audit or clean sources in the main OpenClaw notebook or any of its 4 sub-notebooks, (2) remove duplicate, obsolete, or low-value sources from any OpenClaw NotebookLM notebook, (3) configure or update the AI persona/instructions for any OpenClaw sub-notebook, (4) validate the retrieval quality of OpenClaw knowledge bases, (5) do any deep research, pruning, or restructuring task related to the OpenClaw NotebookLM workspace.
license	Apache-2.0
metadata	{"version":"1.0","capabilities":["notebooklm-auditing","openclaw","content-pruning","persona-configuration","deep-research"]}

OpenClaw NotebookLM Auditor

This skill provides a structured workflow for auditing and refining the "OpenClaw: The Universal Personal AI Assistant Platform" NotebookLM workspace.

OpenClaw is a Universal Personal AI Assistant Platform. Its NotebookLM workspace is organized into a main notebook (master index) and 4 specialized sub-notebooks covering distinct functional areas. The goal of this skill is to ensure each notebook is a high-signal, low-noise knowledge base that functions as a specialized expert on its domain.

Notebook Map

Notebook	Focus	Priority Mode
Main	"OpenClaw: The Universal Personal AI Assistant Platform" — Master Index	Reference only
Notebook A	Core Architecture & Systems	Reliability (stable docs, specs, ADRs)
Notebook B	Plugins & Integrations	Deep Research (breadth, compatibility)
Notebook C	Education & Media	Reliability (accuracy, freshness)
Notebook D	Developer Tools	Deep Research (APIs, SDKs, CLI tools)

For detailed sub-notebook personas, source criteria, and test queries, see references/sub-notebooks.md.

---

Operational Workflow

Phase 1: Reconnaissance

1. Use notebook_list to confirm all 5 notebooks are present (Main + A, B, C, D).
2. For the target notebook(s), call source_list to get all sources with metadata (wordCount, tokenCount, title, URL).
3. Read the current configure_chat instructions via notebook_get to understand the existing persona.
4. Ask the user which notebook(s) to audit if not specified (default: all 4 sub-notebooks).

Phase 2: Diagnosis — Pruning Criteria

Apply these criteria to every source. Flag sources that meet one or more:

Category	Criterion	Action
Duplicate	Same content from multiple URLs or near-identical wordCount + title	Delete all but the most complete version
Obsolete	Content dated >12 months unless it is a foundational spec/ADR	Flag for user confirmation before delete
Off-topic	Content not related to the notebook's declared scope (see Notebook Map)	Delete
Low-value	Thin content (<500 words), marketing fluff, or pure promotional pages	Delete
Redundant	A summary article that is already covered in full by a primary source	Delete the summary, keep the primary

Threshold: If content overlap > 80% between two sources, keep only the most complete/authoritative one.

Phase 3: Planning

Generate a prioritized action list before executing any changes:

1. List sources to DELETE with reason.
2. List sources to ADD to fill identified gaps.
3. List changes to configure_chat (persona update).
4. List notes to create as permanent anchors inside the notebook.

Present the plan to the user and get confirmation before Phase 4.

Phase 4: Execution with Checkpoint

1. Checkpoint: Record the full source list and a baseline query response (notebook_query) before any changes.
2. Apply deletions via source_delete (confirmed sources only).
3. Apply additions via source_add or notebook_add_url / notebook_add_text.
4. Update persona via configure_chat using the target template from references/sub-notebooks.md.
5. Create anchor notes via note_create if needed.
6. If quality degrades, restore from checkpoint.

Phase 5: Validation

Run the validation queries from references/sub-notebooks.md for the target notebook. Evaluate:

• Context Relevance: Does the answer use only on-topic sources?
• Coverage: Does the answer reference diverse sources (not just 1-2)?
• Faithfulness: Are claims traceable to actual source content?
• Anti-hallucination: Does the notebook refuse off-topic queries correctly?

Stop when all 3 test queries return high-quality, on-topic answers.

---

Global Principles

• Less is more: A notebook with 15 excellent sources outperforms one with 50 mediocre ones.
• Scope discipline: Each sub-notebook must refuse to answer questions outside its declared scope.
• Never delete without checkpoint: Always capture state before destructive operations.
• User confirmation required for: deleting sources older than 12 months, deleting >5 sources at once, or any source the user marked as important.