inferred from GitHub
claude harness wiki
AI knowledge base maintained by Claude Code as the knowledge engineer — LLM Wiki - a living memory designed to grow
Core trick
Claude Code edits a wikilinked Obsidian vault of 110+ markdown pages, keeping related concepts connected as the wiki grows.
Evidence
README matched ""Cerebral Valley" "Opus 4.6"": raw/ingested/affaan-m-ecc-readme-part0-ecosystem-tools.md
README excerpt
# 🧠 LLM Wiki — AI Knowledge Base > A structured, graph-based knowledge base about AI, Claude Code, and agentic systems — maintained by Claude Code itself. > **110+ pages in French** | Built with Obsidian | Powered by Claude --- ## 🌍 Language note All wiki pages are written in **French**. You do not need to speak French to use this knowledge base: - Open any page in **Chrome**, click the three dots (⋮) → **Translate to English** - Or paste the content into any AI and ask it to translate The AI-centric usage patterns described below work in any language. --- ## 🤔 What is this? This is a personal knowledge base inspired by [Andrej Karpathy's LLM Wiki pattern](https://karpathy.ai), built and maintained with **Claude Code** as the knowledge engineer. It covers: - 🤖 **Agentic AI systems** — harnesses, agents, orchestration - 🧩 **Claude Code** — hooks, commands, plugins, memory, optimization - 🔐 **Agentic security** — CVEs, sandboxing, prompt injection - 🏗️ **Harness engineering** — pipeline design, session continuity, verification - 🧪 **Evaluation methodology** — graders, pass@k, ablation studies - 🎨 **AI creative tools** — design agents, video generation --- ## 🗂️ Structure ``` llm-wiki/ ├── wiki/ ← 110+ knowledge pages (the core) │ ├── index.md ← table of contents — start here │ ├── templates/ ← ready-to-use harness templates │ └── *.md ← one file = one atomic concept ├── raw/ │ └── ingested/