# AI Engineering System

Machine-readable rules for Cursor / Claude Code / other agents.  
**Source of truth for engineering constraints.** Prompts are helpers, not the core.

## Priority (what AI depends on most)

1. `constitution.md` — laws
2. `architecture.md` — layers
3. `domain.md` — ubiquitous language
4. `patterns/` — how to implement
5. `examples/` — copy shapes
6. `playbooks/` — step flows
7. `review.md` + `definition-of-done.md` + `checklists/`
8. `prompts/` — optional task templates

Also always: `ai-contract.md`, `forbidden.md`, `file-map.md`, `workflow.md`, `commands.md`, `adr/`.

## Tree

```
.ai/
├── constitution.md
├── architecture.md
├── tech-stack.md
├── domain.md
├── domain/
│   └── domain-map.md     # DDD contexts / aggregates
├── product/
│   ├── feature-map.md    # 愈心谷 Feature Tree
│   └── cece-feature-map.md
├── file-map.md
├── workflow.md
├── ai-contract.md
├── coding.md
├── api.md
├── database.md
├── ui.md                 # IA pointer → design/
├── design/               # AI Design System Contract
│   ├── design-system.md
│   ├── component-catalog.md
│   └── platform/
├── deployment.md
├── testing.md
├── security.md
├── review.md
├── definition-of-done.md
├── forbidden.md
├── commands.md
├── adr/
├── patterns/
├── playbooks/
├── checklists/
├── examples/
└── prompts/          # auxiliary
```

## Cursor

`.cursor/rules|templates|commands` are **symlinks** into this directory.  
Truth = `.ai/` only. See `../.cursor/README.md`.

## Evolution

See `ROADMAP.md` — Phase 2/3 (anti-patterns, metrics, gates, memory, graph, mcp, skills) are deferred on purpose.

Root entry: `AGENTS.md` · `CLAUDE.md`