Turn your Obsidian vault into an AI identity layer. Any AI agent reads
ME.md+AGENT.mdand instantly knows who you are, how you think, and how to work with you — built on Knowledge Palace v2 (PARA + Johnny.Decimal + Zettelkasten + MOC + LLM Wiki + Memory Palace).
中文说明：README.zh-CN.md
Every new chat, every new project, every new AI tool — you start from zero. You re-explain your tech stack, your communication style, your preferences. Personal API ends that loop:
ME.md— your single-page identity contractAGENT.md— your behavior contract- A vault navigation layer — for when AI needs depth
Read ME.md once → AI is calibrated for the entire session.
Agent-agnostic by design. This is a folder convention plus two markdown contracts. It doesn't depend on any specific AI runtime — Claude Code, Codex, Cursor, ChatGPT, Gemini, or your custom agent all read the same files the same way.
# 1. Set your vault path
export OBSIDIAN_VAULT_PATH="/path/to/your/vault"
# 2. Run the scaffolder (full Knowledge Palace v2 structure)
bash scripts/setup.sh
# 3. Open your vault and fill in the [PLACEHOLDER]sWant a lighter footprint? Use bash scripts/setup.sh --minimal — only creates the identity layer, skips the 30.knowledge/ tree.
| File | Role |
|---|---|
templates/ME.md |
Identity contract — your "About Me" page for AI |
templates/AGENT.md |
Behavior contract — language, tone, output format, tool rules |
templates/methodology.md |
Knowledge management operating manual (placed at 30.knowledge/00.system/) |
scripts/setup.sh |
One-command vault scaffolder (full or --minimal) |
SKILL.md |
Skill manifest with metadata, examples, use cases |
ME.md is the entry point. The vault splits into two tracks: Track A — Identity Archive (human-curated, read-only for AI) and Track B — Knowledge Production (AI-led with human review).
The contract: Identity is yours (Track A is read-only for AI). Knowledge production scales with AI (Track B is fair game for AI to compile, link, archive).
The Knowledge Production track is a lifecycle pipeline. Material flows in from the left, gets routed through rooms based on its current life stage, and exits as an output or archive.
Core formula: Folders solve lifecycle. MOCs solve topic membership. Wikilinks solve relationships.
Lifecycle:
capture → intelligence/research → notes → frameworks/projects → outputs → archive
- New project onboarding — drop AI into a fresh repo, it reads
ME.mdand matches your style immediately - Cross-tool consistency — Claude Code, Codex, Cursor, ChatGPT, Gemini all use the same identity contract
- Persistent context — stop re-explaining your preferences every session
- Identity-grounded outputs — AI writes/decides in your voice, not generic boilerplate
- Knowledge-grounded decisions — AI references your accumulated notes instead of hallucinating
Personal API is the entry point to a knowledge architecture that fuses six well-established methodologies:
| Method | Contribution |
|---|---|
| PARA (Tiago Forte) | Lifecycle-based directory layout |
| Johnny.Decimal | Numbered prefixes for stable locations |
| Zettelkasten (Luhmann) | Atomic permanent notes |
| MOC / LYT (Nick Milo) | Semantic maps over deep folders |
| LLM Wiki (Karpathy) | Strict raw vs compiled separation |
| Memory Palace | Spatial metaphor for low-cost lookup |
See SKILL.md for the full architecture, AI operation boundaries, frontmatter spec, FAQ, and maintenance routines.
Tested with:
- Claude Code — auto-loads
CLAUDE.mdat vault root for hard rules - Codex / OpenAI Agents — auto-loads
AGENTS.mdat vault root - Cursor — point its rules at
ME.md+AGENT.md - ChatGPT / Gemini / Custom LLM — paste the standard prompt:
"Read ME.md and AGENT.md to understand my context."
The protocol is just markdown. Any agent that can read files can use this skill.
Your filled-in ME.md and AGENT.md contain personal context. Do not commit them to public repositories. A .gitignore is included to help. This skill ships only templates — never your data.
personal-knowledge-vault— cross-project entry skill that pulls context from this vaultknowledge-palace-builder— full step-by-step vault construction guide
Designed and battle-tested daily by @beiyuii. Methodology synthesizes work from Tiago Forte, Niklas Luhmann, Nick Milo, Andrej Karpathy, and Johnny Decimal.
License: MIT.