Read in Korean (νκ΅μ΄λ‘ 보기)
AIWF is an AI-powered workflow framework that enables autonomous software development with Claude Code. It provides intelligent command sets and workflow definitions that allow AI to manage entire project lifecycles - from planning to deployment - with minimal human intervention.
This project is an updated version of Simone.
AIWF is built as a Claude Code Plugin following the official Anthropic plugin architecture. The framework is organized into 4 specialized plugins:
aiwf/
βββ .claude-plugin/
β βββ marketplace.json # Plugin registry
βββ plugins/
βββ aiwf-core/ (26 files) β Session, YOLO, Sprint/Task management
βββ aiwf-dev/ (19 files) β TDD, DevOps automation
βββ aiwf-experts/ (24 files) β Expert agents (architects, reviewers)
βββ aiwf-tools/ (51 files) β Development tools
| Plugin | Description | Key Features |
|---|---|---|
| aiwf-core | Core framework | Session management, YOLO mode, Sprint/Task execution |
| aiwf-dev | Development workflow | TDD (red/green/refactor), DevOps, CI/CD |
| aiwf-experts | Expert agents | Build fixers, architects, code reviewers |
| aiwf-tools | Utility tools | Refactoring, debugging, API scaffolding, docs |
Multi-agent architecture where AIWF acts as the master orchestrator managing entire development workflows, while Claude Code operates as specialized subagents handling specific tasks. YOLO mode enables completely autonomous development - AI handles planning, coding, testing, and deployment without human intervention. Built-in checkpoint system ensures resilience and recovery.
Workflow-based state management with dependency tracking, priority calculation, and smart task recommendations. AI understands project context and suggests optimal next actions.
Five expert personas (Architect, Security, Frontend, Backend, Data Analyst) with specialized knowledge bases and context-aware compression for domain-specific optimization.
Intelligent context construction and optimization that maximizes AI effectiveness while minimizing token usage. Dynamic context compression, workflow-aware prompt engineering, and automatic relevance filtering ensure AI always has the right information at the right time.
Token-efficient context strategies with automatic compression, intelligent chunking, and relevance scoring. Maintains comprehensive project understanding while optimizing for AI model limits and performance.
AIWF operates as the master orchestration agent that coordinates complex development workflows. Claude Code instances function as specialized subagents, each handling specific domains like coding, testing, or documentation. This distributed approach enables parallel processing, fault tolerance, and specialized expertise application.
Native support for Claude Code, Cursor, Windsurf, GitHub, and Git with automated hooks, issue creation, and PR management.
Advanced installation validation with intelligent file checking, dependency verification, and automated troubleshooting. Enhanced validation architecture ensures reliable framework setup and operation.
Install AIWF as a Claude Code plugin directly:
# From GitHub repository
claude plugin add https://github.com/moonklabs/aiwf
# Or from local path (for development)
claude plugin add /path/to/aiwfAfter installation, verify the plugins are loaded:
claude plugin listYou should see 4 AIWF plugins:
aiwf-core- Session, YOLO, Sprint/Taskaiwf-dev- TDD, DevOpsaiwf-experts- Expert agentsaiwf-tools- Development tools
npm install -g aiwfAfter global installation, navigate to your project directory and run:
aiwf installThe installer will guide you through:
- Language Selection: Choose between English and Korean
- Project Setup: Initialize AIWF in your current directory
- Plugin Installation: Install Claude Code plugins
- Documentation: Download guides and templates
- Validation: Comprehensive installation verification
# Option A: Claude Code Plugin (Recommended)
claude plugin add https://github.com/moonklabs/aiwf
# Option B: NPM + Manual Setup
npm install -g aiwf
cd your-project
aiwf installEnter Plan Mode - describe what you want to build and your project goals. Do not start coding yet.
- Engage in thorough Q&A to clarify objectives
- Define project scope and requirements
- Set clear success criteria
# Create/update Claude configuration
/init
# Initialize AIWF framework (creates first milestone)
/aiwf:initialize# Generate sprints from milestone (first sprint becomes active)
/aiwf:create-sprint M01
# If only one milestone exists, you can omit the milestone code
/aiwf:create-sprintReview generated plans and refine through Q&A until satisfied with the approach.
# Create detailed tasks for active sprint (first task becomes active)
/aiwf:create-sprint-tasks# Execute individual task
/aiwf:do-task [task-id]
# OR run autonomous development for entire sprint
/aiwf:yoloYOLO Mode orchestrates multiple Claude Code subagents to execute all sprint tasks continuously without interruption. The master agent coordinates planning while specialized subagents handle coding, testing, documentation, and commits automatically.
aiwf-core (Core Framework)
/aiwf:initialize- Initial project setup/aiwf:session-start- Start a new session/aiwf:session-end- End current session with summary/aiwf:yolo- Autonomous development mode/aiwf:do-task- Execute individual task/aiwf:create-sprint- Create sprint from milestone/aiwf:create-sprint-tasks- Generate tasks for sprint/aiwf:commit- Smart commit with context
aiwf-dev (Development Workflow)
/tdd:tdd-red- Write failing test first/tdd:tdd-green- Implement to pass test/tdd:tdd-refactor- Refactor with tests passing/devops:deploy-checklist- Deployment verification/devops:docker-optimize- Docker optimization
aiwf-tools (Utilities)
/refactoring:refactor-clean- Clean code refactoring/debugging:error-trace- Error tracing and analysis/api:api-scaffold- API scaffolding/docs:doc-generate- Documentation generation
Revolutionary multi-agent orchestration that enables completely autonomous development workflows. AIWF acts as the master agent coordinating multiple Claude Code subagents:
# Full autonomous development - AI handles everything
/aiwf:yolo
# Target specific sprint execution
/aiwf:yolo S03
# Continuous multi-sprint development
/aiwf:yolo sprint-allMaster agent analyzes requirements and coordinates specialized subagents that create tasks, write code, run tests, and commit changes - achieving true multi-agent autonomous software development.
AIWF provides several specialized CLI tools for different aspects of project management:
aiwf install- Install AIWF framework in current projectaiwf create-project- Create new project from templatesaiwf compress- Context compression for token optimizationaiwf token- Token usage monitoring and managementaiwf persona- AI persona managementaiwf evaluate- AI response and code quality evaluation
aiwf-sprint create- Create new independent sprintaiwf-sprint list- List all sprintsaiwf-sprint status- Show sprint status and progressaiwf-sprint task create- Add tasks to sprintaiwf-sprint task execute- Execute sprint tasks
aiwf-checkpoint create- Create project checkpoint for rollbackaiwf-checkpoint list- List all available checkpointsaiwf-checkpoint restore- Restore project to previous checkpoint
aiwf-lang status- Check current language settingsaiwf-lang set ko- Switch to Koreanaiwf-lang set en- Switch to English
aiwf-cache download- Download templates to local cacheaiwf-cache list- List all cached templatesaiwf-cache clean- Clean cache storageaiwf-cache update- Update cached templatesaiwf-cache status- Check cache status and usage
your_project/
βββ .aiwf/ # Project management root
β βββ 00_PROJECT_MANIFEST.md # Central tracking document
β βββ 01_PROJECT_DOCS/ # Project documentation
β βββ 02_REQUIREMENTS/ # Milestone requirements
β βββ 03_SPRINTS/ # Sprint execution tracking
β βββ 04_GENERAL_TASKS/ # Standalone tasks
β βββ 05_ARCHITECTURAL_DECISIONS/ # ADR documents
β βββ 10_STATE_OF_PROJECT/ # Project state snapshots
β βββ 98_PROMPTS/ # Useful AI prompts
β βββ 99_TEMPLATES/ # Document templates
βββ .claude-plugin/ # Plugin registry
β βββ marketplace.json # 4 plugins registered
βββ plugins/ # Plugin content
β βββ aiwf-core/ # Core: session, yolo, sprint
β βββ aiwf-dev/ # Dev: TDD, DevOps
β βββ aiwf-experts/ # Expert agents
β βββ aiwf-tools/ # Development tools
βββ .cursor/rules/ # Cursor IDE development rules
βββ .windsurf/rules/ # Windsurf IDE development rules
Based on your language selection (aiwf-lang set en/ko):
| Language | Commands | Documentation |
|---|---|---|
| English (default) | Standard commands | English docs & templates |
| Korean | _kr suffix available |
Korean docs & templates |
AIWF features a significantly improved validation system that ensures reliable installation and operation:
- 86% Code Reduction: Streamlined validator.js from 348 lines to 48 lines
- Unified Interface: Single
validateInstallation()function replaces 3 redundant methods - Constants-Based Configuration: All validation parameters centralized in
VALIDATION_CONSTANTS - Enhanced Error Reporting: Specific, actionable error messages with detailed diagnostics
const VALIDATION_CONSTANTS = {
MIN_FILE_SIZE: 10, // Minimum file size requirement
MIN_RULE_FILE_SIZE: 50, // Minimum size for AI tool rule files
MIN_FILE_COUNT: {
CURSOR_MDC: 2, // Required .mdc files for Cursor
WINDSURF_MD: 2, // Required .md files for Windsurf
CLAUDE_COMMANDS: 4 // Required command files for Claude
}
};- Multi-Tool Support: Validates Claude Code, Cursor, Windsurf, and Gemini CLI
- File Integrity Checks: Size validation, accessibility verification, and structure validation
- Intelligent Error Recovery: Detailed troubleshooting guidance for failed validations
- Performance Optimized: Faster execution with reduced memory footprint
The codebase follows clean architecture principles for maintainability and extensibility:
- Unified Validation Logic: Consolidated validation functions eliminate redundancy
- Shared Constants: Centralized configuration reduces maintenance overhead
- Streamlined Error Handling: Consistent error reporting across all validation types
- Clear Separation of Concerns: Validation, file management, and error reporting are properly isolated
- Maintainable Code Structure: Reduced complexity makes the codebase easier to understand and modify
- Performance Benefits: Optimized code paths improve installation speed and reliability
- Commands Guide - Complete list of AIWF commands
- CLI Usage Guide - Detailed CLI tool documentation
- Getting Started - Quick start guide for new users
- Development Guide - Contributing to AIWF
- Independent Sprint Guide - YOLO-focused sprint creation
- Checkpoint System Guide - Recovery and progress tracking
- AI Personas Guide - Using specialized AI personas
- Context Compression Guide - Token optimization strategies
- Feature Git Integration Guide - Git hooks and tracking
- Architecture - System architecture and design
- State Management Guide - Workflow-based state system
- AI Workflow - AI integration patterns
- API Reference - Programmatic usage
- Complete API Reference - Comprehensive API documentation
- Validator API Reference - Validation system API documentation
- Code Cleanup Guide - Code cleanup principles and patterns
- Troubleshooting - Common issues and solutions
This installer fetches the AIWF framework from: https://github.com/moonklabs/aiwf
MIT