TrueHOOHA

TrueHOOHA / LLM-Wiki-Skilled

Public

This is my implementation of Karpathy's LLM Wiki. A key highlight is the use of agent skills to enforce workflow rigidity and mitigate agent behavioral deviation.

gist.githubusercontent.comkarpathy442a6bf555914893e9891c11519de94frawac46de1ad27f92b28ac95459c782c07f6b8c964allm-wiki.md
16
2
85% credibility
Found May 23, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

LLM Wiki is a personal knowledge management system where you (the human) collect sources like articles, books, and notes, while an AI assistant handles all the organizational work. The wiki is just a folder of text files you can browse in a special app called Obsidian, which shows you how everything connects. When you add a new source, the AI reads it, creates summaries, links related ideas together, and updates your index. When you have a question, the AI searches your wiki and synthesizes an answer with citations. You can also run health checks to find broken links or outdated information. The project includes tools to verify everything is working correctly and follows a philosophy that humans should curate sources and ask questions, while AI handles the tedious bookkeeping.

How It Works

1
💡 You discover a smarter way to manage knowledge

You learn about keeping your notes, articles, and ideas all organized by an AI assistant instead of doing it all yourself.

2
📁 You set up your personal wiki

You create a folder on your computer where all your knowledge will live - a place you can browse and explore anytime.

3
📄 You drop in your first source

You save an article, book chapter, or podcast notes into your wiki's special inbox folder - just like dropping a file into a shared drive.

4
🤖 Your AI assistant jumps in to help

You ask your AI assistant to process the new source, and it reads it, creates summaries, links related ideas together, and updates your index - all automatically.

5
🔍 You explore and ask questions

You browse your growing wiki using a special app that shows you how everything connects. You ask your assistant 'what do we know about this topic?' and it finds and combines the answers for you.

6
✨ Your knowledge base stays healthy

You run a quick check and your assistant scans for broken links, outdated information, and missing connections - then fixes them with your approval.

🎉 Your knowledge grows and compounds

Every article you add, every question you ask, and every insight you capture builds on everything else - your personal wiki becomes smarter over time.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is LLM-Wiki-Skilled?

This is a Python-based system that turns Obsidian into an AI-powered personal knowledge base. The core idea: you drop sources into a folder, and an LLM agent handles all the tedious maintenance -- summarizing articles, creating linked pages, updating cross-references, and logging changes. Instead of maintaining a wiki yourself, you direct the agent and browse the results. The project enforces strict workflows through "skills" that keep the LLM on track and prevent it from drifting off-script.

Why is it gaining traction?

The killer feature is workflow rigidity. Most wiki agents go off the rails or skip steps. This system uses explicit skill definitions that the agent must follow, catching deviations before they compound into messy wikis. It also ships with practical tooling: schema linting to catch broken links and missing frontmatter, index rebuilding from page metadata, and log validation to ensure append-only history. You get verification tests out of the box, so the system can self-diagnose whether workflows are actually working.

Who should use this?

Researchers building deep-dive wikis on papers and articles. Developers who want a personal knowledge base that compounds over time without maintenance burden. Teams running internal wikis fed by meeting notes and Slack threads. Writers who want a chapter-by-chapter companion wiki while reading a book. Anyone frustrated that their second brain always falls apart because maintaining links is tedious.

Verdict

This is a clever pattern at a very early stage -- 16 stars and thin real-world usage evidence. The verification infrastructure and schema enforcement are solid foundations, but the credibility score of 0.85% reflects limited community validation. Worth experimenting with if you want to offload wiki maintenance to an agent, but treat it as a learning project rather than production infrastructure until adoption grows.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.