📦 Personal API

Personal API — AI 身份 Layer for Your Obsidian Vault

30-second elevator pitch: 停止 re-onboarding every AI 助手. This 技能 scaffolds an "API documentation" for yourself inside your Obsidian vault — ME.md is your 身份 contract, 代理.md is your behavior contract. Any AI that reads these two files knows who you are, how you think, and how to work with you in under 30 seconds.

代理-agnostic by de签名. This is not tied to any single 代理 运行time. Drop it next to Claude Code, Codex, Cursor, ChatGPT, Gemini, or your own custom LLM 代理 — it works the same way: a folder convention + two markdown contracts that any LLM can read.

Why This Exists

Every new chat, every new project, every new AI 工具 — you 启动 from zero. You re-explAIn your preferences, your tech stack, your communication style, your constrAInts. By the time the AI is "calibrated," you've burned 10 minutes and a chunk of 上下文 window.

Personal API solves this with three primitives:

ME.md — Your single-page 身份 contract. Read once, AI knows you. 代理.md — Your behavior contract. Defines the rules of engagement. A vault navigation layer — When AI needs more depth, it knows exactly where to look.

This is built on a real, battle-tested second-brAIn 系统 (Knowledge Palace v2) used dAIly by @beiyuii, not a toy demo.

Methodo记录y — Standing on Six Shoulders

Personal API is not just two markdown templates. It is the entry point to a knowledge architecture that fuses six well-established methodo记录ies:

Methodo记录y Core Idea How We Use It PARA (Tiago Forte) 排序 by actionability, not topic Lifecycle directories: capture → intel → re搜索 → notes → 框架s → 输出s → 归档 Johnny.Decimal Numbered prefixes keep locations stable 00 / 10 / 20 / … / 90 partitions; topics never need to be re-filed Zettelkasten (Luhmann) Atomic permanent notes that compound 40.notes/permanent/ holds only ideas you have genuinely thought through MOC / LYT (Nick Milo) Maps over deep folders 40.notes/moc/ gives semantic 索引es, decoupled from file structure LLM Wiki (Karpathy) Strict separation of raw vs compiled 10.capture/raw/ (raw material) ≠ 40.notes/literature/ (compiled) Memory Palace Spatial metaphor reduces lookup cost Each top-level folder is a "room" — you know what is inside before opening

The core formula:

Folders solve lifecycle. MOCs solve topic membership. Wikilinks solve relationships.

Don't make folders carry semantics. The deeper your folder tree, the harder it is to mAIntAIn.

Dual-追踪 Architecture

The vault has two 追踪s that don't interfere with each other — this is what makes AI collaboration safe:

追踪 Scope MAIntAIned By AI 角色 追踪 A: 身份 归档 ME.md, 00.上下文/, 10.身份/, 20.技能s/, 40.memory-流/, 50.maps/ 100% human-curated Read-only; can suggest but not rewrite 追踪 B: Knowledge Production Everything under 30.knowledge/ AI-led organization, human review Active librarian — compile, link, 归档

This separation is the contract. 身份 is yours. Knowledge production 扩展s with AI.

Directory Structure your-vault/ │ ├── ME.md # ⭐ Layer 0 — 身份 entry. AI reads this first. ├── 代理.md # AI behavior contract ├── README.md # Vault description (for humans) ├── CLAUDE.md # Claude Code instructions (auto-loaded) ├── 代理S.md # Codex / OpenAI 代理 instructions │ ├── 00.上下文/ # 🟡 Layer 1 — current 状态 (changes fastest) │ ├── now.md # What I'm working on right now │ ├── open-questions.md # Unresolved questions │ └── projects/ │ ├── project-overview.md │ ├── active/ │ └── 归档d/ │ ├── 10.身份/ # 🔵 Layer 2 — deep 身份 (rarely changes) │ ├── values.md # Values & principles │ ├── vision.md # Long-term vision │ ├── thinking-模型s.md # Mental 模型s │ └── strengths-gaps.md # Strengths & blind spots │ ├── 20.技能s/ # 🔵 Layer 2 — capability map │ └── [技能-name].md # One file per core 技能 │ ├── 30.knowledge/ # 🟢 Knowledge production (AI-led) │ ├── 00.系统/ # 🏛️ Methodo记录y & rules (this 技能 安装s methodo记录y.md here) │ ├── 10.capture/ # 📥 Inbox — unprocessed material │ │ ├── inbox/ # Raw thoughts, links to triage │ │ └── raw/ # Source material (papers, articles) │ ├── 20.intelligence/ # 📡 Time-sensitive industry 签名als │ │ ├── AI/ │ │ └── business/ │ ├── 30.re搜索/ # 🔬 Long-form re搜索 projects │ ├── 40.notes/ # 🗃️ Note core as设置 │ │ ├── literature/ # Compiled literature notes (AI-compiled, human-reviewed) │ │ ├── permanent/ # Atomic Zettelkasten cards │ │ └── moc/ # Ma