deepinit

Deep Init — Hierarchical Documentation Generation

Analyzes the codebase and generates comprehensive, structured documentation.

Workflow

Step 1: Choose Documentation Scope

Use AskUserQuestion:

{
  "questions": [
    {
      "question": "What documentation do you need?",
      "header": "Docs Type",
      "options": [
        {"label": "README (Recommended)", "description": "Project overview, setup, usage, and contribution guide"},
        {"label": "Architecture", "description": "System design, component diagrams, data flow documentation"},
        {"label": "API docs", "description": "Endpoint specifications with request/response examples"},
        {"label": "Full suite", "description": "README + Architecture + API docs + CHANGELOG"}
      ],
      "multiSelect": false
    }
  ]
}

Step 2: Analyze Codebase

Dispatch explorer (Haiku) to:

• Map complete project structure
• Identify tech stack, frameworks, and tools
• Find existing documentation (README, docs/, comments)
• Detect entry points, routes, models, and key abstractions

Step 3: Generate Documentation

Based on scope, dispatch writer (Haiku) with exploration results:

README: Project overview, installation, quickstart, configuration, features, contributing Architecture: Component diagram (Mermaid), data flow, key decisions, directory guide API docs: Endpoints, parameters, response schemas, examples, error codes CHANGELOG: Parse git log to generate structured changelog

For architecture diagrams, generate Mermaid syntax:

graph TB
    subgraph Frontend
        A[UI Components] --> B[State Management]
    end
    subgraph Backend
        C[API Routes] --> D[Services]
        D --> E[Database]
    end
    B --> C

Step 4: Review and Polish

If "Full suite" was selected, dispatch code-reviewer (Sonnet) to verify:

• Documentation accuracy (do code references match actual code?)
• Completeness (are all major features documented?)
• Clarity (would a new developer understand this?)