š AB Method - Incremental Task Management System
> A revolutionary approach to software development that transforms problems into focused, incremental missions using Claude Code's specialized subagents.
Overview
š AB Method - Incremental Task Management System
A revolutionary approach to software development that transforms problems into focused, incremental missions using Claude Code's specialized subagents.
šÆ Quick Installation
Install AB Method in your project with a single command:
npx ab-method
This will:
- ā
Install all AB Method workflows in
.ab-method/ - ā
Add 11 slash commands to Claude Code (both
/ab-masterand individual commands) - ā
Create necessary directories (
docs/architecture/,tasks/) - ā
Set up or update
CLAUDE.mdwith instructions - ā Install builtin specialized agents for enhanced development workflow
After installation, open Claude Code and choose your preferred approach:
Quick Access (NEW!):
/create-task # Create new tasks with technical details
/create-mission # Transform tasks into focused missions
/extend-task # Add new missions to existing tasks
/analyze-project # Complete project analysis
# ... and 7 more individual commands
Traditional Approach:
/ab-master # Master controller for all workflows
That's it! You're ready to start using the AB Method with enhanced command access.
š Table of Contents
- Overview
- Core Philosophy
- System Architecture
- Getting Started
- Workflow Files Documentation
- Usage Examples
- Task Lifecycle
- Mission Execution Flow
- Architecture Documentation
- Advanced Features
šÆ Overview
The AB Method is an incremental task management system designed specifically for Claude Code that breaks down complex software development projects into manageable tasks and missions. Each mission is completed entirely before moving to the next, building knowledge incrementally while maintaining focus and conserving context.
š§ Core Philosophy
Key Principles:
- One Task at a Time - Maintain laser focus on a single task
- Incremental Missions - Each mission builds on previous knowledge
- Backend First - For full-stack tasks, start with backend to provide types and data
- Validation Checkpoints - User validates before implementation begins
- Continuous Documentation - Architecture and progress tracked in real-time
Why AB Method?
- Context Conservation - Avoid overwhelming the model with too much planning
- Knowledge Building - Each mission uses learnings from previous ones
- No Redundancy - Prevent implementing the same thing twice
- Clear Progress - Always know exactly where you are
šļø System Architecture
.claude/
āāā commands/
āāā ab-master.md # Traditional master controller
āāā create-task.md # Direct task creation
āāā resume-task.md # Resume existing tasks
āāā create-mission.md # Direct mission creation
āāā resume-mission.md # Resume missions
āāā test-mission.md # Create comprehensive tests
āāā extend-task.md # Add missions to existing tasks
āāā analyze-project.md # Full project analysis
āāā analyze-frontend.md # Frontend analysis
āāā analyze-backend.md # Backend analysis
āāā update-architecture.md # Architecture updates
.ab-method/
āāā core/ # Core workflow files
ā āāā analyze-project.md # Project analysis orchestrator
ā āāā analyze-frontend.md # Frontend architecture analysis
ā āāā analyze-backend.md # Backend architecture analysis
ā āāā update-architecture.md # Architecture documentation updater
ā āāā create-task.md # Task creation workflow
ā āāā resume-task.md # Task resumption workflow
ā āāā create-mission.md # Mission creation workflow
ā āāā resume-mission.md # Mission resumption workflow
ā āāā extend-task.md # Task extension workflow
ā
āāā utils/ # Mission utility files
ā āāā backend-mission.md # Backend mission coordinator
ā āāā frontend-mission.md # Frontend mission coordinator
ā āāā planning-mission.md # Planning mission coordinator
ā
āāā structure/ # Configuration
ā āāā index.yaml # Paths and structure configuration
docs/
āāā architecture/ # Generated architecture docs
āāā tech-stack.md
āāā entry-points.md
āāā frontend-patterns.md
āāā backend-patterns.md
āāā external-services.md
āāā project-constraints.md
tasks/ # Created tasks and missions
āāā [task-name]/
āāā progress-tracker.md
āāā mission-*.md
š¦ Getting Started
Two Ways to Access Workflows
š Quick Access (Recommended for experienced users):
/create-task # Create new task with enhanced technical details
/resume-task # Resume paused tasks
/create-mission # Transform tasks into missions
/resume-mission # Continue incomplete missions
/test-mission # Create comprehensive tests
/extend-task # Add new missions to existing tasks
/analyze-project # Complete project analysis
/analyze-frontend # Frontend architecture analysis
/analyze-backend # Backend services analysis
/update-architecture # Maintain architecture docs
š Traditional Controller (Great for beginners):
/ab-master [workflow-name]
The master controller provides:
- View all available workflows (call without arguments)
- Start specific workflows with guidance (provide workflow name)
- Help text and workflow descriptions
Available Workflows
| Workflow | Purpose | When to Use |
|---|---|---|
analyze-project | Full project architecture analysis | Starting a new project or onboarding |
analyze-frontend | Deep frontend analysis | Understanding client-side architecture |
analyze-backend | Deep backend analysis | Understanding server-side architecture |
create-task | Create a new development task | Starting new feature/fix |
resume-task | Resume an existing task | Continuing previous work |
create-mission | Create next mission for a task | Ready for next step |
resume-mission | Resume an in-progress mission | Continuing mission work |
extend-task | Add new missions to existing task | Requirements changed or scope expanded |
update-architecture | Update architecture docs | After implementing features |
š Workflow Files Documentation
Core Workflows
š analyze-project.md
Purpose: Orchestrates comprehensive project analysis using specialized subagents.
Flow:
- Prompts user to confirm parallel agent deployment
- Deploys frontend-developer and backend-architect agents simultaneously
- Creates comprehensive architecture documentation
Usage:
/ab-master analyze-project
š± analyze-frontend.md & š¾ analyze-backend.md
Purpose: Deep-dive analysis of frontend/backend architecture.
Key Features:
- Reads from
.ab-method/structure/index.yamlfor output paths - Creates detailed patterns documentation
- Identifies tech stack and conventions
š create-task.md
Purpose: Creates a new task with all missions defined upfront.
Task Status Flow:
Brainstormed ā Validated ā In dev ā Testing ā Completed
Process:
- Define problem statement, context, and expected outcome
- Identify task type (Frontend/Backend/Full-stack)
- Create all missions upfront
- Initialize with "Brainstormed" status
š resume-task.md & resume-mission.md
Purpose: Resume work from exact breakpoint.
Features:
- Shows current progress visually
- Identifies last action and next step
- Maintains continuity across sessions
Utility Files
š§ backend-mission.md
Responsibilities:
- Loads backend architecture documentation
- Coordinates backend-architect and backend-developer agents
- Ensures DTOs use database types for maintainability
šØ frontend-mission.md
Responsibilities:
- Loads frontend patterns and backend types
- Coordinates UX expert and frontend-developer agents
- Integrates with backend missions for type safety
š planning-mission.md
Responsibilities:
- Loads ALL architecture documentation
- Handles research and design decisions
- Documents rationale and trade-offs
š” Usage Examples
Example 1: Starting a New Feature (Quick Style)
# 1. Create task directly with enhanced technical context
/create-task
# 2. System asks for problem definition + technical details
"Create a todos table that fetches from API and displays in frontend"
"Follow existing shadcn/ui patterns, use our current API structure"
"Testing with vitest, ensure responsive design"
# 3. Creates task with comprehensive technical context:
# Technical Context: Code constraints, architecture hints, tech requirements
# Code Guidance: File organization, testing requirements, performance
# - Mission 1: Backend - Create todo model and API
# - Mission 2: Frontend - Create table component
# - Mission 3: Integration - Connect frontend to backend
# 4. System prompts for validation
"Task created with status 'Brainstormed'. Ready to validate?"
# 5. After validation, start first mission
/create-mission # Deploys nextjs-backend-architect agent
Example 1b: Starting a New Feature (Traditional Style)
# 1. Use master controller for guidance
/ab-master create-task
# 2. Same enhanced flow but with help text and explanations
# 3. Creates same comprehensive task documentation
# 4. Use /ab-master create-mission for guided mission creation
Example 2: Resuming Work
Quick Style:
# Resume directly
/resume-task
> "Which task to resume?"
> "todo-table"
# Shows progress with agent tracking:
# ā Mission 1: Backend API - COMPLETED (nextjs-backend-architect)
# ā³ Mission 2: Frontend Table - IN PROGRESS (shadcn-ui-adapter)
# Last: Created base component with shadcn/ui patterns
# Next: Add state management and data fetching
# Agent Output: Component documented in docs/tasks/todo-table/
Traditional Style:
# Resume with guidance
/ab-master resume-task
# Same progress display with helpful explanations
š Task Lifecycle
graph LR
A[Problem Definition] --> B[Create Task]
B --> C[Define All Missions]
C --> D[Brainstormed Status]
D --> E[User Validation]
E --> F[Validated Status]
F --> G[Mission 1]
G --> H[Mission 2]
H --> I[Mission N]
I --> J[Task Complete]
šÆ Mission Execution Flow
graph TB
A[Create Mission] --> B{Mission Type?}
B -->|Backend| C[Load backend-mission.md]
B -->|Frontend| D[Load frontend-mission.md]
B -->|Planning| E[Load planning-mission.md]
C --> F[Backend Architect Agent]
F --> G[Backend Developer Agent]
D --> H[UX Expert Agent]
H --> I[Frontend Developer Agent]
E --> J[Research/Planning Agent]
G --> K[Update Progress]
I --> K
J --> K
K --> L[Mission Complete]
š Architecture Documentation
The system automatically generates and maintains architecture documentation:
Generated Files:
- tech-stack.md - Technologies, frameworks, and tools
- entry-points.md - API endpoints and application entries
- frontend-patterns.md - Component architecture and patterns
- backend-patterns.md - API design and service patterns
- external-services.md - Third-party integrations
- project-constraints.md - Limitations and requirements
Update Strategy:
After implementing features, run:
/ab-master update-architecture
This will:
- Assess impact of changes
- Update relevant documentation
- Preserve history with timestamps
- Mark deprecated features
š Advanced Features
Specialized Sub-Agent System
The AB Method leverages Claude Code's specialized agents for enhanced development:
Built-in Agents (Installed by default):
- shadcn-ui-adapter - UI component creation and styling
- nextjs-backend-architect - Next.js backend development
- sst-cloud-architect - Serverless infrastructure
- vitest-component-tester - Component testing
- playwright-e2e-tester - End-to-end testing
- ascii-ui-mockup-generator - UI mockups and wireframes
- mastra-ai-agent-builder - AI agent development
- qa-code-auditor - Code quality analysis
Agent Coordination Flow:
graph TB
A[Mission Start] --> B{Mission Type?}
B -->|Backend| C[nextjs-backend-architect]
B -->|Frontend| D[shadcn-ui-adapter]
B -->|Testing| E[vitest-component-tester]
C --> F[Document in docs/]
D --> F
E --> F
F --> G[Update Progress Tracker]
Comprehensive Documentation Output
All agent work is automatically documented in structured locations:
Architecture Documentation:
docs/architecture/- Generated by analysis workflows- Technical constraints, patterns, and tech stack details
- Updated continuously as missions complete
Task Documentation:
docs/tasks/[task-name]/- Individual task folders- Progress trackers with technical context sections
- Mission-specific documentation with agent outputs
Agent Output Tracking: Each mission tracks which agents were used and their contributions:
## Agent Usage Tracking
### Mission 1 Agents
- nextjs-backend-architect: Created API endpoints and data models
- qa-code-auditor: Performed code quality analysis
## Sub-Agent Outputs
### Backend Architecture Plan (nextjs-backend-architect)
- Database schema: users, todos tables
- API endpoints: GET/POST /api/todos
- Type definitions: TodosTable, UserTable
Enhanced Technical Context (NEW!)
Create-task workflow now includes comprehensive technical guidance:
Technical Context Sections:
- Code Constraints - File naming, coding standards, patterns
- Architecture Hints - Services to reuse, integration points
- Tech Stack Requirements - Required libraries, versions, dependencies
- API Constraints - Endpoint naming, authentication patterns
Code Guidance Sections:
- File Organization - Directory structure, import patterns
- Testing Requirements - Coverage expectations, test frameworks
- Performance Considerations - Caching, optimization requirements
Parallel Agent Execution
The analyze-project workflow deploys multiple specialized agents in parallel for maximum efficiency:
- Frontend Expert Agent
- Backend Architect Agent
Type Safety Across Stack
Backend missions generate types that frontend missions automatically use:
// Backend creates:
interface Todo {
id: string;
title: string;
status: TodosTable["status"]; // Database type
}
// Frontend uses:
const TodoList: React.FC<{ todos: Todo[] }> = ...
Incremental Knowledge Building
Each mission document contains:
- Dependencies from previous missions
- Files created/modi