LLM Wiki — Second Brain for Claude Code + Obsidian — Development

LLM Wiki turns Claude Code into a disciplined maintainer of a persistent, interlinked Obsidian knowledge base (second brain). It incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current, so knowledge compounds across sessions instead of being re-derived by RAG on every query.

engineering claude-code

📂 View skill on GitHub ↗ 🚀 How to use

✨ What it does

Ingests sources into a persistent wiki, creating source summaries and updating 10-15 relevant pages.
Maintains cross-references and an index.md catalog, updated on every ingest.
Runs health checks (lint) to flag contradictions, stale claims, orphan pages, and missing cross-refs.
Answers queries using an index-first search and synthesizes with citations, offering to file good answers back.
Provides slash commands (/wiki-init, /wiki-ingest, /wiki-query, /wiki-lint, /wiki-log) and sub-agents for delegated work
Ships Python scripts (stdlib only) for vault init, ingest, index update, logging, search, lint, and graph analysis.

🎯 When to use it

Building a personal knowledge base for goals, health, psychology, or journaling.
Conducting long-term research on a topic, ingesting papers, articles, and reports over weeks.
Creating a companion wiki for a book, filing chapters and tracking characters, themes, and plot threads.
Maintaining an internal business or team wiki fed by Slack, meeting notes, and calls.
Any scenario where you want knowledge to accumulate across sessions rather than being re-derived by RAG.

🚀 How to use

Trigger: /wiki-ingest

Trigger with slash commands like /wiki-ingest, /wiki-query, /wiki-lint, or by asking to 'build a second brain' or 'ingest this paper'. Initialize a vault with /wiki-init or the init script, then drop sources into raw/ and ingest them. Provide file paths or questions as inputs.

Example prompts:

/wiki-ingest ~/vaults/research/raw/anthropic-monosemanticity.pdf
/wiki-query "how does monosemanticity compare to mechanistic interpretability?"

📄 Output: A maintained Obsidian vault with raw sources, wiki pages (entities, concepts, sources, comparisons, synthesis), an index, and a log.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/engineering/llm-wiki/skills/llm-wiki .claude/skills/llm-wiki

Skill source: engineering/llm-wiki/skills/llm-wiki/SKILL.md

⚠️ Good to know

Requires Obsidian as the vault environment and is not suited for one-shot Q&A over fixed documents; the LLM never edits raw/ files, so source corrections must be done manually and re-ingested.

❓ FAQ

What is the iron rule of the LLM Wiki skill?

The LLM never edits files in raw/; sources are immutable. All LLM writes go to wiki/.

How does the wiki scale beyond hundreds of pages?

At ~100 sources or hundreds of pages, index.md and filesystem search suffice; beyond that, layer in a local search tool like qmd or use scripts/wiki_search.py.

Can this skill be used with tools other than Claude Code?

Yes, the schema is stored in AGENTS.md for Codex, Cursor, Antigravity, OpenCode, and Gemini CLI, and the Python scripts are standard-library only, so they run everywhere.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.