Claude Code Source Study

# Claude Code Source Study

> Skill by [ara.so](https://ara.so) — Daily 2026 Skills collection.

A 25-article deep-dive into Claude Code's ~1900-file source code, covering System Prompt engineering, multi-agent orchestration, tool systems, permission security, and terminal UI. Learn production-grade AI agent patterns from Anthropic's real CLI product.

## What This Project Is

This is a **Chinese-language source code analysis series** that dissects Claude Code (Anthropic's AI CLI coding assistant) module by module — with exact file references, line numbers, and code snippets. Each article extracts reusable design patterns for building your own AI agent applications.

**Tech stack covered:** Bun + TypeScript + Ink (React for terminals) + Anthropic API

## Repository Structure

```
claude-code-source-study/
├── docs/
│   ├── 00-目录与阅读指引.md      # Index and reading guide
│   ├── 01-项目全景.md             # Project overview
│   ├── 02-启动优化.md             # Startup optimization
│   ├── 03-状态管理.md             # State management
│   ├── 04-System-Prompt-工程.md  # System prompt engineering
│   ├── 05-对话循环.md             # Conversation loop
│   ├── 06-上下文管理.md           # Context management
│   ├── 07-Prompt-Cache.md        # Prompt caching
│   ├── 08-Thinking-与推理控制.md  # Thinking & reasoning
│   ├── 09-工具系统设计.md          # Tool system design
│   ├── 10-BashTool-深度剖析.md    # BashTool deep dive
│   ├── 11-命令系统.md             # Command system
│   ├── 12-Agent-系统.md           # Agent system
│   ├── 13-内置Agent设计模式.md    # Built-in agent patterns
│   ├── 14-任务系统.md             # Task system
│   ├── 15-MCP-协议实现.md         # MCP protocol
│   ├── 16-权限系统.md             # Permission system
│   ├── 17-Settings-系统.md       # Settings system
│   ├── 18-Hooks系统.md            # Hooks system
│   ├── 19-Feature-Flag与编译期优化.md
│   ├── 20-API调用与错误恢复.md    # API retry/recovery
│   ├── 21-Ink框架深度定制.md      # Ink UI customization
│   ├── 22-设计系统.md             # Design system
│   ├── 23-Memory系统.md           # Memory system
│   ├── 24-Skill-Plugin开发实战.md # Plugin development
│   └── 25-架构模式总结.md         # Architecture patterns summary
└── README.md
```

## Reading Routes

### ⚡ Quick Route (7 articles) — Global understanding
```
01 → 02 → 03 → 05 → 09 → 12 → 25
```

### 🤖 AI Engineering Route (9 articles) — Deep AI core
```
01 → 03 → 04 → 05 → 06 → 08 → 09 → 12 → 13
```

### 📚 Complete Route (25 articles)
Read docs/01 through docs/25 in order.

## Key Patterns Extracted from Claude Code

### 1. Tool Builder Pattern (`buildTool()`)

Claude Code registers tools using a builder with three-layer conditional registration:

```typescript
// Pattern extracted from docs/09-工具系统设计.md
const buildTool = <TInput, TOutput>(config: {
  name: string
  description: string
  inputSchema: ZodSchema<TInput>
  handler: (input: TInput, context: ToolContext) => Promise<TOutput>
  isEnabled?: (context: AppContext) => boolean
  requiresPermission?: PermissionLevel
}) => config

// Registration with conditions
const tools = [
  buildTool({ name: 'bash', ... }),
  buildTool({ name: 'read_file', ... }),
  buildTool({ name: 'write_file', ... }),
].filter(tool => tool.isEnabled?.(ctx) ?? true)
```

### 2. AsyncGenerator Conversation Loop (`docs/05`)

```typescript
// Pattern: state-machine conversation loop using AsyncGenerator
async function* conversationLoop(
  messages: Message[],
  tools: Tool[]
): AsyncGenerator<StreamEvent> {
  while (true) {
    const stream = await anthropic.messages.stream({
      model: 'claude-opus-4-5',
      messages,