Summary

This project provides an experimental agent skill for OpenClaw/Codex designed to build persistent, LLM-compiled knowledge bases. It shifts from reactive RAG retrieval to proactive knowledge synthesis, allowing information to compound over time, with the LLM managing writing and cross-referencing.

How It Works

The core approach involves an LLM compiling raw sourced documents into a structured, cross-linked Markdown wiki. Unlike RAG, which re-retrieves documents per query, this method synthesizes knowledge into a persistent, evolving resource. The LLM handles all writing, cross-referencing, and bookkeeping, enabling knowledge to compound and become richer with each interaction.

Quick Start & Requirements

• Installation: Copy the skill directory into your agent's skills path (e.g., cp -r llm-wiki/ ~/.claude/skills/llm-wiki/).
• Prerequisites: Python 3.x, Node.js (for companion tools).
• Scaffold New Wiki: python3 llm-wiki/scripts/scaffold.py ~/my-wiki "My Research Topic"
• Add Source: Copy Markdown files to ~/my-wiki/raw/articles/.
• Ingest Source: Instruct agent: "ingest raw/articles/my-article.md"
• Query Wiki: Instruct agent: "what does the wiki say about X?"
• Lint Wiki: python3 llm-wiki/scripts/lint_wiki.py ~/my-wiki
• Web Viewer Setup: Requires npm install and npm run build in audit-shared and web directories.
• Web Viewer Start: cd web && npm start -- --wiki "/path/to/your/wiki-root" --port 4175
• Obsidian Plugin Setup: Requires npm install, npm run build, npm run link in plugins/obsidian-audit, then enable in Obsidian.

Highlighted Details

• Includes an Obsidian plugin (plugins/obsidian-audit/) for in-vault text selection and audit filing.
• Features a local Node.js preview server (web/) rendering wikis with Mermaid, KaTeX, and wikilinks, supporting feedback submission.
• A shared TypeScript library (audit-shared/) ensures byte-identical audit file structures between Obsidian and the web viewer.
• The lint_wiki.py script performs a 7-pass health check on the wiki's integrity.

Maintenance & Community

Feedback is primarily collected via GitHub issues. The project is inspired by Andrej Karpathy's llm-wiki Gist and related community efforts.

Licensing & Compatibility

The project is released under the MIT license, which is permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

This skill is experimental and subject to iteration and change over time.