skill
Golden-Second-Brain
An Autonomous, Multi-Agent LLM Wiki & Second Brain Template for Obsidian. Inspired by Karpathy's LLM Wiki pattern.
Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/daffaavicenna/Golden-Second-Brain itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 25, 2026🧠 GOLDEN SECOND BRAIN
An Autonomous, Multi-Agent LLM Wiki & Second Brain Template for Obsidian.
Inspired by Andrej Karpathy's LLM Wiki Pattern.
💡 The Philosophy
Most AI note-taking systems rely on traditional RAG: you upload documents, retrieve chunks, and the model re-derives insights from scratch every single time. Nothing accumulates.
Golden Second Brain changes this paradigm:
- Obsidian is the IDE.
- The LLM Agent is the software engineer.
- The Wiki is the persistent, compounding codebase.
Instead of recalculating insights on every prompt, the LLM continuously reads incoming documents from an inbox/, synthesizes them into an interlinked web of Markdown pages (wiki/), updates a central catalog (index.md), and moves raw sources to archive/ so they are never re-processed redundantly.
✨ Key Features
- Universal Multi-Agent Ready: Works seamlessly with Claude Code, OpenAI Codex, Google Antigravity, Nous Hermes, OpenClaw, OpenCode, Cursor, Aider, Grok, and custom bots via standardized
AGENTS.md&SYSTEM.md. - Zero Redundant Processing: Raw documents staged in
inbox/are ingested and automatically moved toarchive/YYYY-MM/. - Bilingual Efficiency Pattern:
- Wiki Storage (100% English): Maximum token efficiency (saving 40–60% tokens) and superior semantic density for the AI.
- User Interaction (Bahasa Indonesia / User's Native Language): AI communicates, explains, and interviews the user in their preferred language.
- Built-in Obsidian Skills: Native support for Obsidian Flavored Markdown, wikilinks (
[[Page Name]]), callouts, tags, Canvas (.canvas), and Bases (.base). - Interactive Onboarding: New users are guided through an onboarding interview to build
User Profile.md, giving any AI agent instant context without repetitive introductory prompts. - Automatic Version Control: Automated Git commits track every ingestion step (
git commit -m "ingest: <Title>").
📁 Repository Structure
Golden Second Brain/
├── inbox/ # 📥 STAGING: Drop raw files here (.md, .pdf, .txt, webclips)
│ └── README.md
├── archive/ # 📦 ARCHIVE: Immutable storage of ingested files
│ └── YYYY-MM/ # (e.g. archive/2026-10/)
│
├── wiki/ # 🧠 PERSISTENT WIKI (Maintained by LLM)
│ ├── concepts/ # Core theories, mental models, patterns, frameworks
│ ├── entities/ # Specific people, software, organizations, User Profile
│ ├── summaries/ # Structural extractions and takeaways of source documents
│ └── synthesis/ # Cross-domain analyses, deep comparisons, query answers
│
├── index.md # 📖 Master catalog of all wiki pages
├── log.md # 📜 Append-only chronological timeline of vault operations
│
├── .agents/ # 🤖 Agent skills (obsidian-markdown, obsidian-cli, defuddle, etc.)
├── AGENTS.md # ⚙️ Master multi-agent configuration
├── SYSTEM.md # ⚙️ Universal system directive for Hermes, OpenClaw, and tool agents
├── PROMPTS.md # 💬 Copy-paste prompts for web chatbots (Grok, ChatGPT, Claude Web)
├── CLAUDE.md # ⚙️ Configuration for Claude Code
├── GEMINI.md # ⚙️ Configuration for Gemini / Antigravity
├── .cursorrules # ⚙️ Configuration for Cursor / Windsurf
└── README.md # 📘 Documentation
🚀 Quick Start Guide
1. Clone & Open
- Clone this repository:
git clone https://github.com/your-username/golden-second-brain.git - Open the cloned folder as a Vault in Obsidian.
2. Launch Any AI Agent
Launch your preferred AI tool inside the vault root:
- Claude Code:
claude - Codex / OpenCode:
opencode/codex - Antigravity / Gemini CLI: Open workspace in Antigravity
3. Complete Onboarding
On your first session, the agent will detect an uninitialized profile in wiki/entities/User Profile.md and conduct a quick onboarding interview in Bahasa Indonesia to configure your personal goals, tech stack, and active projects.
4. Ingest Your First Document
- Drop any note, paper, or web clip into
inbox/. - Tell your agent:
"Tolong ingest file yang ada di inbox"
- Watch the agent compile the knowledge into
wiki/, updateindex.md, log the action inlog.md, and move the raw file intoarchive/.
🛠️ Operations Reference
| Operation | Command / Prompt | What Happens |
|---|---|---|
| Ingest | ingest inbox / "Tolong ingest" | Reads raw files from inbox/, updates wiki/, moves raw files to archive/YYYY-MM/, updates index.md & log.md, commits git. |
| Query | "Jelaskan tentang X berdasarkan wiki" | Searches index.md, reads relevant wiki pages, delivers synthesized response. High-value answers are saved to wiki/synthesis/. |
| Lint | lint wiki / "Cek kesehatan wiki" | Scans for orphan pages, missing concept notes, claim contradictions, and catalog integrity. |
📜 License
MIT License. Created with ❤️ for knowledge explorers and PKM enthusiasts.