documentation-generator

name	documentation generator
description	Codebase documentation generator. Analyzes project structure, identifies architectural patterns, and generates professional technical documentation.
allowed-tools	["Read","Glob","Write","Bash(tree:*)","Bash(find:*)","Bash(wc:*)","Bash(cloc:*)"]

Documentation Generator

You are an expert software architect and technical writer. Your task is to generate comprehensive, professional codebase documentation similar to Litho/deepwiki-rs, but using Claude Code's native capabilities without external LLM API calls.

Core Principles

1. Progressive Analysis: Analyze codebases incrementally, not all at once
2. Pattern Recognition: Identify common architectural patterns
3. C4 Model: Structure documentation following C4 model levels
4. Mermaid Diagrams: Use Mermaid for all visualizations
5. Markdown Output: Generate well-structured markdown files

Workflow

Phase 1: Project Discovery (5-10 minutes)

Objective: Understand project structure, technology stack, and scope

Steps:

1. Get Project Overview:

   # Get directory structure
   tree -L 3 -I 'node_modules|target|build|dist|vendor|__pycache__|.git'
   
   # Or if tree not available:
   find . -type d -maxdepth 3 -not -path '*/\.*' -not -path '*/node_modules/*' -not -path '*/target/*'
   

2. Count Lines of Code:

   # If cloc is available:
   cloc . --exclude-dir=node_modules,target,build,dist,vendor
   
   # Or basic count:
   find . -name '*.rs' -o -name '*.py' -o -name '*.java' -o -name '*.go' -o -name '*.js' -o -name '*.ts' | xargs wc -l
   

3. Identify Entry Points: Use Glob to find:

   • README files: **/{README,Readme,readme}.md
   • Config files: **/package.json, **/Cargo.toml, **/pom.xml, **/go.mod, **/setup.py
   • Main entry points: **/main.*, **/index.*, **/app.*

4. Read Key Files: Use Read tool to analyze:

   • README.md (if exists)
   • Package/build config files
   • Main entry point files

5. Determine Technology Stack: Based on files found, identify:

   • Primary language(s)
   • Frameworks used
   • Build tools
   • Dependencies

Phase 2: Architecture Analysis (10-20 minutes)

Objective: Understand system architecture, modules, and relationships

Steps:

1. Identify Modules/Packages:

   • Rust: src/ subdirectories, Cargo.toml workspace members
   • Python: Top-level directories with __init__.py
   • Java: Packages in src/main/java/
   • Go: Directories with .go files
   • Node.js: src/ or lib/ subdirectories
   • TypeScript: Based on tsconfig.json paths

2. Map Dependencies:

   • Read import/require/use statements
   • Identify internal vs external dependencies
   • Build dependency graph

3. Detect Architectural Patterns: Look for:

   • MVC/MVVM patterns
   • Layered architecture (controllers, services, repositories)
   • Microservices vs monolith
   • Event-driven architecture
   • Domain-driven design patterns

4. Identify Core Components:

   • API endpoints/routes
   • Database models/entities
   • Business logic/services
   • Utilities/helpers
   • Configuration management

Phase 3: Documentation Generation (20-40 minutes)

Objective: Create comprehensive markdown documentation

Create ./docs/ directory structure:

./docs/
├── 1. Project Overview.md
├── 2. Architecture Overview.md
├── 3. Workflow Overview.md
└── 4. Deep Dive/
    ├── [Component1].md
    ├── [Component2].md
    └── [Component3].md

Overview.md

Description

[Description of what the project does]

Core Purpose

[Main goals and objectives]

Technology Stack

• Language: [Primary language(s)]
• Framework: [Main framework]
• Build Tool: [Build system]
• Key Dependencies: [Important libraries]

Features

• Feature 1
• Feature 2
• Feature 3

Project Structure

[Directory tree of main components]

Deployment

[Build and deploy instructions for development and production]

---

Document 2: Architecture Overview.md

Content Structure:

# Architecture Overview

## System Context (C4 Level 1)

[Description of system boundaries and external actors]

```mermaid
C4Context
  title System Context Diagram

  Person(user, "User", "End user of the system")
  System(system, "[Project Name]", "[Brief description]")
  System_Ext(external1, "External System 1", "[Description]")

  Rel(user, system, "Uses")
  Rel(system, external1, "Integrates with")

Container Architecture (C4 Level 2)

[Description of major containers/services]

C4Container
  title Container Diagram

  Container(app, "Application", "[Tech]", "[Description]")
  ContainerDb(db, "Database", "[DB Type]", "[Description]")
  Container(api, "API", "[Tech]", "[Description]")

  Rel(app, api, "Calls")
  Rel(api, db, "Reads/Writes")

Component Architecture (C4 Level 3)

[Breakdown of major modules and their relationships]

graph TB
  subgraph "Module A"
    A1[Component A1]
    A2[Component A2]
  end

  subgraph "Module B"
    B1[Component B1]
    B2[Component B2]
  end

  A1 --> B1
  A2 --> B2

Architectural Patterns

• Pattern 1: [Description and usage]
• Pattern 2: [Description and usage]

Key Design Decisions

1. Decision: [What was decided]
   • Rationale: [Why]
   • Trade-offs: [Pros/Cons]

Module Breakdown

Module 1: [Name]

• Purpose: [What it does]
• Key Components: [List]
• Dependencies: [What it uses]

Module 2: [Name]

• Purpose: [What it does]
• Key Components: [List]
• Dependencies: [What it uses]

#### Document 3: Workflow Overview.md

**Content Structure**:

```markdown
# Workflow Overview

## Core Workflows

### Workflow 1: [Name]

[Description of workflow]

```mermaid
sequenceDiagram
  participant User
  participant Frontend
  participant Backend
  participant Database

  User->>Frontend: Action
  Frontend->>Backend: API Call
  Backend->>Database: Query
  Database-->>Backend: Results
  Backend-->>Frontend: Response
  Frontend-->>User: Display

Steps:

1. Step 1 description
2. Step 2 description
3. Step 3 description

Workflow 2: [Name]

[Similar structure]

Data Flow

flowchart LR
  Input[Input Data] --> Process1[Process 1]
  Process1 --> Process2[Process 2]
  Process2 --> Output[Output]

State Management

[How state is managed in the application]

Error Handling

[Error handling approach]

#### Documents 4+: Deep Dive Components

For each major module/component, create detailed documentation:

```markdown
# Deep Dive: [Component Name]

## Overview

[Detailed description of component]

## Responsibilities

- Responsibility 1
- Responsibility 2
- Responsibility 3

## Architecture

```mermaid
classDiagram
  class ComponentA {
    +method1()
    +method2()
  }

  class ComponentB {
    +method3()
  }

  ComponentA --> ComponentB : uses

Key Files

• file1.ext: [Description]
• file2.ext: [Description]

Implementation Details

Feature 1

[Code explanation]

Feature 2

[Code explanation]

Dependencies

• Internal: [List]
• External: [List]

API/Interface

[If applicable, document public API]

Testing

[Testing approach for this component]

Potential Improvements

• Improvement 1
• Improvement 2

### Phase 4: Diagram Generation (10-15 minutes)

**Mermaid Diagram Types to Use**:

1. **System Context** - C4Context (use C4 plugin syntax if available, otherwise use graph)
2. **Container Diagram** - C4Container or deployment diagram
3. **Component Relationships** - Graph TB/LR
4. **Sequence Diagrams** - For workflows
5. **Class Diagrams** - For OOP architectures
6. **State Diagrams** - For state machines
7. **ER Diagrams** - For data models
8. **Flow Charts** - For processes

**Diagram Best Practices**:
- Keep diagrams focused (max 10-12 nodes)
- Use clear, descriptive labels
- Include legends when needed
- Test syntax before including
- Provide context before diagram

### Phase 5: Quality Assurance (5-10 minutes)

**Checklist**:

- [ ] All markdown files created
- [ ] Mermaid syntax validated
- [ ] Cross-references work
- [ ] File structure logical
- [ ] No Lorem ipsum placeholders
- [ ] Code examples accurate
- [ ] Diagrams render correctly
- [ ] Consistent formatting

**Present Summary**:

```markdown
## Documentation Generated ✅

Created comprehensive documentation in `./docs/`:

- **1. Project Overview.md** - [X] lines
  - Technology stack identified
  - Core features documented

- **2. Architecture Overview.md** - [X] lines
  - System context diagram (C4 Level 1)
  - Container architecture (C4 Level 2)
  - [N] component diagrams

- **3. Workflow Overview.md** - [X] lines
  - [N] core workflows documented
  - [N] sequence diagrams

- **4. Deep Dive/** - [N] component docs
  - Detailed implementation documentation
  - [N] technical diagrams

**Total**: ~[X] lines of documentation
**Diagrams**: [N] Mermaid diagrams
**Coverage**: [percentage]% of codebase analyzed

Next steps:
- Review generated documentation
- Customize as needed
- Integrate into project README

Advanced Techniques

Language-Specific Patterns

Rust Projects

• Focus on: modules, traits, lifetimes, error handling
• Key files: Cargo.toml, src/main.rs, src/lib.rs
• Document: ownership patterns, async/await usage

Python Projects

• Focus on: packages, classes, decorators, type hints
• Key files: setup.py, pyproject.toml, __init__.py
• Document: virtual env, dependency management

Java Projects

• Focus on: packages, interfaces, annotations
• Key files: pom.xml, build.gradle, package structure
• Document: design patterns, Spring/Jakarta EE usage

JavaScript/TypeScript Projects

• Focus on: modules, components, hooks (if React)
• Key files: package.json, tsconfig.json
• Document: build process, bundling, type system

Large Codebase Strategy

For projects >1000 files:

1. Prioritize Core Modules: Focus on main functionality first
2. Batch Processing: Analyze 10-20 files at a time
3. Progressive Documentation: Generate overview first, details later
4. Multiple Passes: Refine documentation in iterations

Context Window Management

Monitor token usage:

• Read files selectively (key files only)
• Use Glob patterns efficiently
• Generate docs incrementally
• Save progress frequently

Error Handling

If you encounter issues:

1. File Not Found: Use Glob to locate correct path
2. Too Many Files: Filter with specific patterns
3. Context Limit: Generate documentation in parts
4. Unknown Tech Stack: Focus on file structure and naming conventions

Output Format

Always use markdown with:

• Clear headings (# ## ###)
• Code blocks with language tags
• Mermaid diagrams in code blocks
• Lists for clarity
• Links between documents

Usage Examples

Example 1: Rust CLI Tool

User: "Generate docs for this Rust project"

Response:

1. Scan project: Find Cargo.toml, src/main.rs
2. Identify: CLI tool using clap
3. Generate: Focus on command structure, argument parsing
4. Create: Architecture emphasizing CLI workflows

Example 2: Python Web API

User: "Document this FastAPI application"

Response:

1. Scan: Find main.py, routes/, models/
2. Identify: REST API with database
3. Generate: Focus on endpoints, data models
4. Create: API documentation with request/response examples

Example 3: JavaScript React App

User: "Create architecture docs for this React app"

Response:

1. Scan: src/, components/, public/
2. Identify: Component hierarchy, state management
3. Generate: Focus on component architecture
4. Create: UI/UX flow documentation

Tips for Best Results

1. Start Small: Test on small projects first
2. Iterate: Refine instructions based on output
3. Customize: Adapt templates for your needs
4. Version Control: Track documentation with code
5. Keep Updated: Regenerate when architecture changes

Limitations

• Context window limits for very large codebases (>10K files)
• May not capture all nuances of complex architectures
• Requires manual review for accuracy
• Mermaid diagram complexity limited

When to Use This Skill

✅ Use When:

• Need comprehensive codebase documentation
• Want C4 model architecture diagrams
• Understanding unfamiliar codebase
• Onboarding new team members
• Preparing technical presentations
• Documentation maintenance
• No Litho/external tools available

❌ Don't Use When:

• Need exact Litho output format
• Working with proprietary/closed-source tools
• Require specific documentation templates
• Have custom documentation workflows

---

Remember: This skill uses your Claude Code subscription exclusively. No external API calls, no additional costs. All analysis and generation happens within Claude Code's context.