A persistent wiki system for AI agents to organize raw materials into an interlinked knowledge base compatible with Obsidian.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install llm-kb
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-kb 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 Base (llm-kb)?
The llm-kb skill transforms your AI agent into a sophisticated knowledge base curator. Inspired by Andrej Karpathy's wiki method, it moves beyond temporary file uploads to create a growing, persistent vault of knowledge. By automating the extraction of concepts, entities, and summaries from raw data, this skill ensures that your insights are never lost and remain fully navigable via tools like Obsidian.
Using Openclaw Skills allows developers to maintain a structured second brain where every claim is traceable to its source. The agent doesn't just chat; it continuously refines and links information, ensuring the knowledge base grows in value over time rather than becoming a disorganized pile of files.
LLM Knowledge Base (llm-kb) Use Cases
- Researching complex technical topics and synthesizing multiple sources into a unified wiki.
- Building a personal second brain or technical documentation vault that evolves with your projects.
- Organizing project-specific knowledge to provide long-term memory for AI coding agents.
- Converting unstructured web research and terminal outputs into structured concept and entity pages.
- Performing automated health checks and linting on large markdown knowledge bases to ensure link integrity.
How LLM Knowledge Base (llm-kb) Works
- Initialization: The agent creates a standardized directory structure including raw and wiki folders with pre-configured Obsidian settings.
- Ingestion: Raw materials such as URLs, local files, or pasted text are captured in the inbox and processed into structured source cards.
- Extraction: The curator identifies key concepts and entities, creating or updating dedicated markdown pages with automatic wikilinks.
- Navigation Management: The agent updates the central index, activity log, and overview pages to maintain a clear map of the knowledge vault.
- Querying and Output: Users can query the wiki or request complex deliverables like comparison tables or reading guides based on the existing knowledge.
LLM Knowledge Base (llm-kb) Setup
To initialize a new vault using Openclaw Skills, prepare your environment with the following directory structure:
KB_NAME="my-technical-kb"
mkdir -p "$KB_NAME"/{raw/inbox,raw/assets,wiki/sources,wiki/concepts,wiki/entities,wiki/outputs,.obsidian}
Once the directories are created, the agent generates an AGENTS.md file in the root directory. This file acts as the primary instruction set for any AI agent entering the vault, ensuring consistent maintenance rules and data standards across different LLM platforms.
LLM Knowledge Base (llm-kb) Data Schema & Taxonomy
The skill organizes data into a strict hierarchy to ensure compatibility with markdown editors and graph visualization tools:
| Directory | Content Type | Purpose |
|---|---|---|
raw/inbox/ |
Markdown/Text | Unmodified original source materials |
wiki/sources/ |
Source Cards | Summaries and key points linked to raw data |
wiki/concepts/ |
Concept Pages | Explanations of abstract or technical ideas |
wiki/entities/ |
Entity Pages | Profiles for people, companies, or products |
wiki/outputs/ |
Deliverables | AI-generated analyses, guides, and reports |
wiki/index.md |
Navigation | The entry point for browsing the knowledge base |
LLM Knowledge Base (llm-kb) Advanced Features
- Automatic link integrity checks (LINT) to identify orphan pages, broken wikilinks, or duplicate concepts.
- Cross-agent compatibility through the AGENTS.md standard, allowing Claude, Cursor, or Codex to collaborate on the same vault.
- Support for Marp-compatible markdown slide deck generation directly from wiki content.
- Native Obsidian configuration files included during setup for instant graph view and backlink support.
- Deep-dive research mode that automates web searching and multi-source ingestion for new topics.
SKILL.md
Loading
Related Openclaw Skills
Solscan Pro Solana Data Skill
An advanced tool for AI agents to retrieve professional-grade Solana on-chain data including accounts, tokens, NFTs, and DeFi activities.
README Craft
A professional-grade documentation agent that transforms technical projects into high-conversion open-source repositories through the cognitive funnel design.
ClawPK Marketplace
A decentralized Agent-to-Agent task marketplace for browsing, completing tasks, and earning USDC rewards via the x402 protocol.
Duleheng Consulting: Engineering Expert System
A professional-grade engineering consulting expert system for strategic cost control, contract auditing, and project management decision support.
Deslop
Deslop is an advanced writing refinement tool that removes AI patterns and injects human rhythm into your documents.
API Doc Gen
A developer-centric tool that generates consistent API documentation from code comments and type hints.
Featured*
