LLM-Wiki

Build a long-lived personal knowledge base where an LLM compiles raw sources into linked markdown pages.

Obsidian = IDE, LLM = programmer, wiki = codebase

A distributable skill and starter template for people who want to turn books, articles, notes, and web pages into a maintainable markdown knowledge system instead of a pile of files.

At a glance

• Purpose: build a personal knowledge base that behaves more like a codebase than a notebook pile
• Core workflow: collect - extract - recon - ingest - query - lint
• Output: linked source / concept / entity / comparison / chapter / decision pages
• Reading layer: Obsidian
• Execution model: any coding agent that can run terminal commands and write files (Claude Code, Codex, OpenCode, …)

Why this workflow exists

Most note systems stop at collection. Most RAG systems stop at retrieval.

LLM Wiki is for incremental knowledge maintenance:

• raw files stay in raw/
• extraction artifacts stay in raw/extracted/
• durable knowledge pages live in wiki/
• Obsidian becomes the navigation layer, not the authoring engine

What you get

• a reusable, GitHub-installable skill
• a starter repository skeleton for new personal knowledge bases
• a page model for source, concept, entity, comparison, chapter, and decision pages
• source import guidance for pdf, html, md, txt, docx, and epub
• Obsidian setup guidance for navigation, dashboards, and visual mapping