Understanding Existing Code with AI DevKit

When joining a new project or working with unfamiliar code, understanding how everything fits together can be overwhelming. The document-code skill analyzes your codebase from any entry point and generates comprehensive documentation with visual diagrams—helping you understand complex systems in minutes instead of hours.

Common scenarios:

• You just joined a team and need to understand the authentication system
• You're debugging an issue and need to trace the complete execution flow
• You're planning a refactor and need to see all dependencies

Before using document-code, ensure you have:

• Initialized AI DevKit in your project (ai-devkit init)
• An AI editor with skill support

Note: Use this directly in your AI editor's chat window, not in the terminal.

Use the document-code skill to analyze <entry-point>

Use the document-code skill to analyze <entry-point>

You can analyze code starting from different types of entry points:

Use the document-code skill to analyze src/api/users.ts

Analyzes a specific file and its dependencies.

Use the document-code skill to analyze src/services/

Analyzes an entire module or directory structure.

Use the document-code skill to analyze calculateTotalPrice

Analyzes a specific function and its call chain.

Use the document-code skill to analyze POST:/api/users

Analyzes complete API request/response flow.

• Natural language descriptions of how code works
• Implementation details and design patterns
• Logic flow and component relationships

• Automatically traces all dependencies
• Maps complete execution paths
• Identifies external integrations

• Flowcharts showing execution paths
• Sequence diagrams for API flows
• Architecture diagrams for modules
• Component relationship maps

• Performance considerations
• Security implications
• Potential improvements
• Refactoring opportunities

When you ask the agent to document src/services/auth/, the AI generates documentation like this:

docs/ai/knowledge/auth-service.md:

# Auth Service Analysis

## Overview
The auth service handles user authentication using JWT tokens...

## Architecture Diagram

flowchart TD
    A[Login Request] --> B[Validate Credentials]
    B --> C{Valid?}
    C -->|Yes| D[Generate JWT]
    C -->|No| E[Return 401]

## Key Components
- `AuthController` — Handles HTTP endpoints
- `TokenService` — Manages JWT creation and validation
- `UserRepository` — Database access for user data

Generated documentation is saved to docs/ai/knowledge/ in your project:

docs/ai/knowledge/
├── auth-service.md       # Analysis of auth module
├── user-api.md           # Analysis of user endpoints
└── diagrams/             # Generated visual diagrams

These files can be committed to version control, making your knowledge base searchable and shareable with your team.

Use Case	Description
Onboarding	Help new developers understand complex systems quickly
Documentation	Generate comprehensive system documentation automatically
Debugging	Understand complete execution flows before diving into code
Refactoring	Get full context and dependency maps before making changes
Knowledge Base	Create searchable, versioned documentation for your team

Problem	Solution
Command not recognized	Run ai-devkit init first to install commands
Analysis seems incomplete	Try analyzing a smaller entry point first
Diagrams not rendering	Ensure your editor supports Mermaid syntax
Entry point not found	Check the path is relative to your project root

• Debug an issue — Use structured debugging with your captured knowledge for context
• Plan a refactor — Use your analysis to inform design review decisions
• Build your knowledge base — Enable Memory to make your captured knowledge searchable
• Extend capabilities — Install Skills to add more AI capabilities