内容摘录
AIDD Framework
AIDD FrameworkParallel Drive
**The standard framework for AI Driven Development**
AI agents like Claude Code ship features fast. aidd Framework keeps them working, secure, and maintainable.
AI agents generate code that runs but fails at scale. GitClear tracked 211 million lines from 2020 to 2024 and found 8x more code duplication as AI adoption increased. Google's DORA report shows AI adoption correlates with 9% higher bug rates and degraded stability. Agents skip tests, couple modules, duplicate logic, and miss vulnerabilities.
AIDD provides the architecture, test workflows, and specification system that turn AI speed into sustainable velocity so you can ship secure, production-ready, maintainable software, quickly.
Includes:
**AIDD CLI** – project bootstrap and automation
**Agent Runtime** – workflows from product discovery to commit and release
**SudoLang Prompt Language** – typed pseudocode for AI orchestration
**Server Framework** – composable backend for Node and Next.js
**Utilities & Component Library** – common patterns and reusable recipes to accelerate your app development
Table of Contents
<!-- START doctoc generated TOC please keep comment here to allow auto update -->
<!-- DON'T EDIT THIS SECTION, INSTEAD RE-RUN doctoc TO UPDATE -->
About AIDD Framework
🚀 Quick Start with AIDD CLI
📋 Requirements
Detailed Installation Instructions
Development Workflow
🧪 User Testing
Why SudoLang?
What's Included
🚀 AIDD Server Framework
Authentication Middleware
🛠️ AIDD CLI Reference
Installation & Usage
Command Options
Examples
📁 AI System Structure
Key Components
🎯 AI Integration
📋 Vision Document
Why You Need a Vision Document
Creating Your Vision Document
How Agents Use the Vision Document
AGENTS.md File
aidd-custom/ — Project Customization
🔧 Cursor Editor Setup
Automatic Setup (Recommended)
When to Use --cursor
When NOT to Use --cursor
Manual Integration
Troubleshooting
📄 License
🤝 Contributing
<!-- END doctoc generated TOC please keep comment here to allow auto update -->
About AIDD Framework
**AI-Driven Development (AIDD)** is a methodology where AI systems take primary responsibility for generating, testing, and documenting code, automating most of the software creation process so humans can focus on the big picture and 10× their productivity.
AIDD Framework is a collection of reusable metaprograms, agent orchestration systems, and prompt modules that put high-quality software engineering processes on autopilot rails. It implements time-tested workflows including specification-driven development, systematic task planning with Test Driven Development (TDD), and automated code review with best practices enforcement.
**SudoLang** is a pseudocode language for prompting large language models with clear structure, strong typing, and explicit control flow.
**AI Workflow Commands** - Use these in your AI assistant chat (Cursor, ChatGPT, Claude, etc.):
🚀 Quick Start with AIDD CLI
To install for Cursor:
Install without Cursor integration:
📋 Requirements
**Node.js**: 16.0.0+ (requires ESM support)
**Environment**: Unix/Linux shell (bash, zsh) or Windows with WSL
**Editors**: Works with any editor; optimized for Cursor
**LLM**: Works with any sufficiently advanced LLM. As of this writing, we recommend Claude 4.5 Sonnet.
**Agents**: You can ask most agent systems to use this system.
Detailed Installation Instructions
**Install SudoLang syntax highlighting**: Visit the SudoLang Github Repository and install syntax highlighting for your editor.
**Clone the AI system**:
**Create a Vision Document** (important!):
Create a vision.md file in your project root. This document serves as the source of truth for AI agents. See Vision Document for details.
**Explore the structure**:
**Start using AI workflows**:
Reference ai/rules/ in AI prompts for better context
Use ai/commands/ as workflow templates
Customize rules for your specific project needs
This gives you immediate access to:
🤖 **Agent orchestration rules** (ai/rules/)
⚙️ **AI workflow commands** (ai/commands/)
📋 **Development best practices** (JavaScript, TDD, UI/UX)
🎯 **Product management tools** (user stories, journey mapping)
Development Workflow
For features or bug fixes spanning more than a few lines:
**Create a branch**: git checkout -b your-branch-name
**Discover what to build with /discover**: Set up your project profile and discover key user journeys to create a user story map (saved to plan/story-map/)
**Plan execution with /task**: Create a structured epic with clear requirements
**Review with /review**: Eliminate duplication, simplify without losing requirements
**Execute with /execute**: Implement using TDD, one requirement at a time
**Push and PR**: git push origin your-branch-name then open a Pull Request
Note: We use this process to build the aidd framework. See CONTRIBUTING.md for details.
🧪 User Testing
**Validate features with real users and AI agents.** AIDD generates dual testing scripts from user journeys:
**Human scripts** - Think-aloud protocol with video recording for manual testing
**AI agent scripts** - Executable tests with screenshots and persona-based behavior
Research from the Nielsen Norman Group shows that testing with just 3-5 users reveals 65-85% of usability problems. Iterate quickly by testing small, fixing issues, and testing again.
**Quick start**:
📖 **Read the complete User Testing Guide →**
Why SudoLang?
For most simple prompts, natural language is better. Use it. But if you need the AI to follow a program, obey constraints, keep track of complex state, or implement complex algorithms, SudoLang can be extremely useful.
Because of the natural language emphasis, SudoLang is easier to learn than programming languages like JavaScript or Python.
Pseudocode can improve reasoning performance vs natural language prompts, and create shorthands for many prompting styles, such as chain-of-thought reasoning, decision trees, etc.
SudoLang is a declarative, constraint-based, interface…