By kaellim
Knowledge compilation skill for building personal wikis from raw sources — based on Karpathy's LLM Wiki pattern
A Claude Code plugin for building personal knowledge bases using LLMs.
Based on Andrej Karpathy's LLM Wiki pattern (April 2026) — the idea that the correct way to use LLMs with documents is not retrieval (RAG), but knowledge compilation into a persistent, interlinked wiki.
raw/ (you add) → LLM compiles → wiki/ (agent maintains) → Obsidian (you read)
# 1. Add the marketplace
/plugin marketplace add KaelLim/llm-wiki
# 2. Install the plugin
/plugin install llm-wiki@llm-wiki
Copy skills/llm-wiki/SKILL.md content into your agent's system instructions (CLAUDE.md / AGENTS.md).
# 1. Bootstrap a new wiki (creates directories + sets global config)
bash scripts/wiki-bootstrap.sh ~/my-wiki
# 2. Open the wiki directory and start Claude Code
cd ~/my-wiki
claude
# 3. Drop a source and ingest
# (save an article to raw/articles/my-article.md, then:)
> ingest raw/articles/my-article.md
# 4. Query your wiki
> What do I know about distributed systems?
# 5. Health check
> lint
The wiki is a single, central knowledge base — not one per project. The bootstrap script automatically saves the wiki path to ~/.config/llm-wiki/root, so the agent knows where to find it from any project.
To explicitly point a project to the wiki, add this to the project's CLAUDE.md:
wiki-root: /absolute/path/to/my-wiki
The agent will never create raw/ or wiki/ inside a code project directory.
The skill triggers automatically via natural language. Just tell the agent what you want:
> ingest raw/articles/my-article.md ← processes source into wiki pages
> lint ← checks wiki health
> What is event-driven architecture? ← queries wiki, cites pages
All six workflows (ingest, query, lint, enhance, compile, evolve-schema) are documented in skills/llm-wiki/SKILL.md.
Based on Karpathy's three-layer design:
| Layer | Owner | Purpose |
|---|---|---|
raw/ | Human | Immutable source material |
wiki/ | LLM | Compiled, interlinked knowledge pages |
CLAUDE.md | Both | Schema governing wiki conventions |
MIT
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub kaellim/llm-wiki --plugin llm-wikiClaude + Obsidian knowledge companion. Sets up a persistent, compounding wiki vault (Karpathy's LLM Wiki pattern). v1.7 "Compound Vault" + v1.8 methodology modes close 5 of 5 priority gaps from the May 2026 compass artifact. Ships: substrate alignment with kepano/obsidian-skills, default Obsidian CLI transport, hybrid retrieval (contextual prefix + BM25 + cosine rerank per Anthropic's Sept 2024 research), per-file advisory locking for multi-writer safety, pre-commit verifier agent, AND methodology modes (LYT / PARA / Zettelkasten / Generic) for first-class organizational support no other Claude+Obsidian competitor offers. v1.7.x audit closure: every BLOCKER + HIGH + MEDIUM + LOW finding from the v1.7.0 audit is CLOSED or DEFERRED-with-rationale. Optional DragonScale Memory extension (log folds, deterministic addresses, semantic tiling lint, boundary-first autoresearch).
Complete AI coding workflow system. Self-correcting memory + persistent FTS5-indexed research wikis + auto-research loop + multi-LLM council on a single SQLite store. 33 skills, 8 agents, 22 commands, 37 hook scripts across 24 events. Cross-agent via SkillKit.
Build and maintain an LLM-curated personal knowledge base in your project — Andrej Karpathy's LLM Wiki pattern, designed to scale to thousands of pages without becoming a context bottleneck. Now with an optional compiled graph layer for typed, provenance-backed relationships.
Make your AI agent code with your project's architecture, rules, and decisions.