A specialized skill for managing structured Markdown wikis through automated source ingestion, knowledge synthesis, and deterministic maintenance.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install llm-knowledge-bases
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install llm-knowledge-bases using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is LLM Knowledge Bases?
The LLM Knowledge Bases skill allows an AI agent to manage a sophisticated Vault using a dedicated runtime. It transforms raw source materials—including text, PDFs, and images—into a highly structured Markdown wiki. Unlike standard note-taking tools, this skill uses a specific operating model to separate raw assets from compiled source notes, concepts, and entities, ensuring that all information is grounded and verifiable. By integrating with the Openclaw Skills ecosystem, it provides a robust framework for long-term knowledge management and automated research synthesis.
This skill is inspired by expert workflows to ensure that every interaction leaves the wiki better structured. It prioritizes evidence-based writing, automated indexing, and rigorous linting to prevent the accumulation of AI-generated noise. It is an essential tool for developers and researchers who need a reliable, runtime-managed knowledge repository that grows alongside their projects.
LLM Knowledge Bases Use Cases
- Ingesting large volumes of raw text, PDFs, and images into a structured source note library.
- Building a durable repository of concepts and entities derived from multiple research papers.
- Automatically generating cross-source synthesis pages to identify trends or tensions in data.
- Cleaning up placeholder content and fixing stale links within a large Markdown knowledge base.
- Creating archived answer notes that remain grounded in existing wiki evidence.
How LLM Knowledge Bases Works
- The skill starts by running a health check via kb_status to identify missing source notes or structural gaps.
- Raw materials stored in the raw directory are identified and processed based on their file type (text, PDF, or image).
- For multimodal assets, the skill generates representations like OCR or vision notes before compiling the final source note.
- The agent synthesizes information into specific page types including sources, concepts, entities, and syntheses.
- Deterministic tools like kb_lint and kb_repair_source_ids are used to maintain the integrity of the data manifest.
- Finally, the kb_rebuild_indexes tool updates the home index and logs to reflect the latest changes in the wiki.
LLM Knowledge Bases Setup
To enable this skill, ensure your agent environment supports the required Openclaw Skills runtime tools.
# Ensure your directory structure matches the LLM-KB model
mkdir -p vault/raw vault/wiki/sources vault/.llm-kb
# The agent requires access to the following toolset:
# kb_status, kb_lint, kb_prepare_source, kb_upsert_source_note, etc.
Users can initiate workflows using natural language commands or the explicit $llm-knowledge-bases prefix for guaranteed routing.
LLM Knowledge Bases Data Schema & Taxonomy
The skill organizes data into a strict hierarchy within the Vault to maintain separation between raw data and synthesized knowledge:
| Directory | Content Type | Description |
|---|---|---|
raw/ |
Source Assets | Original, unmodified files from the outside world. |
wiki/sources/ |
Source Notes | Compiled Markdown notes grounded in raw files. |
wiki/concepts/ |
Durable Concepts | Pages capturing reusable ideas or frames. |
wiki/entities/ |
Recurrency Entities | Pages for named items that appear across sources. |
wiki/syntheses/ |
High-level Analysis | Cross-source conclusions and thematic synthesis. |
.llm-kb/ |
Runtime State | Internal metadata, OCR, and vision representations. |
LLM Knowledge Bases Advanced Features
- Scenario presets for specific workflows like book-review-batch and ai-topic-cleanup.
- Multimodal support for processing PDFs and images through representation-first paths.
- Deterministic source ID repair to fix manifest drift without manual editing.
- Automated gap mapping to identify repeated ideas that deserve dedicated concept or entity pages.
- Support for bilingual natural-language entry points, optimized for both English and Chinese prompts.
SKILL.md
Loading
Related Openclaw Skills
daily-astro
A bilingual AI-powered daily horoscope skill providing zodiac insights, fortune scores, and automated push notifications for Openclaw Skills users.
Gate News Event Explainer
A specialized tool for explaining the root causes of cryptocurrency price volatility through deep event attribution and on-chain verification.
Daily Sleep Assistant
A personalized sleep coach that automates wind-down routines and provides science-backed sleep hygiene via push notifications.
Daily Idiom (每日成语)
A specialized AI skill for daily Chinese idiom learning, offering historical stories, usage examples, and automated push notifications.
stock-decision
An automated investment assistant for A-share markets that generates real-time trading decisions and position management reports.
Personalized Newsletter Generator (newsletter-init)
An AI-powered information synthesis skill that automates the search, filtering, and delivery of personalized newsletters and research briefs.
Featured*
