Ab Method

Version: 2.3.0 | Repository: GitHub | License: Unknown

Classification

tech: prompt-based exec: sequential function: dev-methodology ecosystem: agnostic scope: project-level integration: drop-in user: solo-dev complexity: moderate maturity: stable community: growing maintenance: active

Scores Summary

SOLID PrinciplesProduction Ready
Overall 3.9/5.0 ⭐⭐⭐☆☆ Overall 74/100 🟡
S - Single Responsibility 4.2/5.0 Reliability 70
O - Open/Closed 3.8/5.0 Observability 75
L - Liskov Substitution 3.7/5.0 Security 75
I - Interface Segregation 4.2/5.0 Performance 82
D - Dependency Inversion 3.8/5.0 Maintainability 70

Full Analysis

AB Method Framework Analysis

Analysis Metadata

  • Date: 2025-11-28
  • Analyst: Claude (via research-framework-analyzer skill)
  • Skill Invocation: CONFIRMED via Skill(“rcr-research:framework-analyzer”)
  • Template Version: 1.1
  • Category: methodology

Analysis Version: 2.3.0 Framework Version: 2.3.0 Previous Version Analyzed: 1.0.0 Category: methodology


Table of Contents

  1. Overview & Context
  2. Architecture & Design
  3. Development Cycle Integration
  4. SOLID Principles Adherence
  5. Production Readiness Assessment
  6. Best Practices & Patterns
  7. Limitations & Trade-offs
  8. Community & Ecosystem
  9. Innovation & Differentiation
  10. References & Sources
  11. Error Handling Patterns
  12. Token Efficiency Analysis M. Methodology Principles

1. Overview & Context

The AB Method is an incremental task management system designed specifically for Claude Code that transforms complex software development projects into focused, sequential missions. Created by Ayoub Bensalah, the framework emerged from the insight that focused, incremental progress produces better results than attempting to tackle entire projects at once.

Since its August 2025 initial release, AB Method has evolved significantly through version 2.3.0, adding intelligent subagent integration (v2.1.0), individual slash commands for all workflows (v2.2.0), and the ability to extend existing tasks with new missions (v2.3.0). The framework targets full-stack development teams, particularly those working with Next.js, shadcn/ui, and SST (Serverless Stack) technologies.

The core innovation lies in mission-based decomposition: work is divided into discrete missions with clear boundaries, each designed to be completed entirely before moving to the next. This approach ensures Claude maintains full context for each piece of work while building progressive knowledge across mission boundaries. The framework reports 60-70% token savings compared to unstructured approaches.

Version 2.3.0 Key Changes (since v1.0.0):

Version Date Change Type Description
v2.0.0 Aug 2025 BREAKING Testing separated into dedicated workflow
v2.1.0 Aug 2025 Feature Intelligent subagent integration, interactive clarification
v2.2.0 Sep 2025 Feature Individual slash commands for all 11 workflows
v2.3.0 Nov 2025 Feature /extend-task command for adding missions to existing tasks

AB Method distinguishes itself through its backend-first philosophy, where types and API contracts are established before frontend work begins. This ensures type safety flows from server to client, preventing the mismatches that commonly plague full-stack projects. The framework includes 8 pre-installed specialized agents for domains ranging from shadcn/ui component integration to end-to-end testing with Playwright.

Installation requires only a single npx ab-method command, making adoption friction minimal. The framework integrates through slash commands (11 total) and maintains project context through a structured CLAUDE.md file that instructs Claude on the methodology for each session.

Sources:


2. Architecture & Design

2.1 Component Architecture

AB Method v2.3.0 implements a hierarchical task structure with clear separation:

┌──────────────────────────────────────────────────────────────┐
│                    AB Method Core v2.3.0                     │
├──────────────────────────────────────────────────────────────┤
│  .ab-method/                                                 │
│  ├── core/          # Workflow executables (10 .md files)    │
│  │   ├── analyze-project.md                                  │
│  │   ├── analyze-frontend.md                                 │
│  │   ├── analyze-backend.md                                  │
│  │   ├── update-architecture.md                              │
│  │   ├── create-task.md                                      │
│  │   ├── resume-task.md                                      │
│  │   ├── create-mission.md                                   │
│  │   ├── resume-mission.md                                   │
│  │   ├── test-mission.md        # NEW in v2.0.0              │
│  │   └── extend-task.md         # NEW in v2.3.0              │
│  ├── utils/         # Mission coordinators                   │
│  │   ├── backend-mission.md                                  │
│  │   ├── frontend-mission.md                                 │
│  │   └── planning-mission.md                                 │
│  └── structure/     # Configuration (index.yaml)             │
├──────────────────────────────────────────────────────────────┤
│  .claude/                                                    │
│  ├── commands/      # 11 slash commands (NEW in v2.2.0)      │
│  │   ├── ab-master.md           # Traditional controller     │
│  │   ├── create-task.md         # Direct task creation       │
│  │   ├── resume-task.md         # Resume paused tasks        │
│  │   ├── create-mission.md      # Transform to missions      │
│  │   ├── resume-mission.md      # Continue missions          │
│  │   ├── test-mission.md        # Testing workflow           │
│  │   ├── extend-task.md         # Add missions (v2.3.0)      │
│  │   ├── analyze-project.md     # Full analysis              │
│  │   ├── analyze-frontend.md    # Frontend analysis          │
│  │   ├── analyze-backend.md     # Backend analysis           │
│  │   └── update-architecture.md # Architecture updates       │
│  └── agents/        # 8 specialized agents                   │
│      ├── shadcn-ui-adapter.md                                │
│      ├── nextjs-backend-architect.md                         │
│      ├── sst-cloud-architect.md                              │
│      ├── vitest-component-tester.md                          │
│      ├── playwright-e2e-tester.md                            │
│      ├── ascii-ui-mockup-generator.md                        │
│      ├── mastra-ai-agent-builder.md                          │
│      └── qa-code-auditor.md                                  │
├──────────────────────────────────────────────────────────────┤
│  docs/architecture/ # Auto-generated documentation           │
│  ├── tech-stack.md                                           │
│  ├── entry-points.md                                         │
│  ├── frontend-patterns.md                                    │
│  ├── backend-patterns.md                                     │
│  ├── external-services.md                                    │
│  └── project-constraints.md                                  │
├──────────────────────────────────────────────────────────────┤
│  tasks/[task-name]/ # Task tracking & missions               │
│  ├── progress-tracker.md                                     │
│  └── mission-*.md                                            │
├──────────────────────────────────────────────────────────────┤
│  CLAUDE.md          # Framework instructions                 │
└──────────────────────────────────────────────────────────────┘

2.2 Dual Access Philosophy (Enhanced in v2.2.0)

The framework provides two invocation paths, significantly improved in v2.2.0:

Quick Style (Recommended for experienced users):

/create-task        # Create new task with technical details
/resume-task        # Resume paused tasks
/create-mission     # Transform tasks into missions
/resume-mission     # Continue incomplete missions
/test-mission       # Create comprehensive tests (NEW in v2.0.0)
/extend-task        # Add missions to existing tasks (NEW in v2.3.0)
/analyze-project    # Complete project analysis
/analyze-frontend   # Frontend architecture analysis
/analyze-backend    # Backend services analysis
/update-architecture # Maintain architecture docs

Traditional Style (Great for beginners):

/ab-master                    # View all workflows
/ab-master create-task        # Guided task creation
/ab-master analyze-project    # Guided analysis

2.3 Specialized Agent System

The framework includes 8 specialized agents deployed based on mission type:

Agent Domain Purpose Deployment Trigger
shadcn-ui-adapter UI shadcn/ui component integration Frontend missions
nextjs-backend-architect Backend Next.js API patterns Backend missions
sst-cloud-architect Infrastructure Serverless deployment Infrastructure missions
vitest-component-tester Testing Component unit tests Test missions
playwright-e2e-tester Testing End-to-end testing E2E test missions
ascii-ui-mockup-generator Design Text-based wireframes Design missions
mastra-ai-agent-builder AI Agent development AI missions
qa-code-auditor Quality Code quality analysis Review missions

Agent Coordination Flow (v2.1.0+):

Mission Start
     │
     ▼
┌──────────────────┐
│ Mission Type     │
│ Detection        │
└────────┬─────────┘
         │
    ┌────┴────┬─────────┬─────────┐
    ▼         ▼         ▼         ▼
Backend   Frontend   Testing   Planning
    │         │         │         │
    ▼         ▼         ▼         ▼
nextjs-   shadcn-   vitest/   Research
backend   ui        playwright Agent
architect adapter   testers
    │         │         │         │
    └────┬────┴─────────┴─────────┘
         ▼
┌──────────────────┐
│ Document Output  │
│ in docs/         │
└────────┬─────────┘
         ▼
┌──────────────────┐
│ Update Progress  │
│ Tracker          │
└──────────────────┘

2.4 Data Flow

Mission execution follows a structured flow with knowledge accumulation:

Task Created
     │
     ▼
┌──────────────────┐
│ Missions Defined │ ← All missions planned upfront
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│ User Validation  │ ← Approval checkpoint
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│  Mission 1 Start │
│  ├── Load Context│ ← Previous + Current
│  ├── Deploy Agents│ ← Specialized agents
│  ├── Execute     │
│  ├── Document    │ ← Auto-generated
│  └── Record Usage│
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│ Mission 1 Complete│
└────────┬─────────┘
         │
    ┌────┴────────────────────┐
    │                         │
    ▼                         ▼
┌──────────────────┐   ┌──────────────────┐
│  Mission 2 Start │   │ /extend-task     │ ← NEW in v2.3.0
│  (accumulated    │   │ Add more missions│
│   knowledge)     │   │ to existing task │
└────────┬─────────┘   └────────┬─────────┘
         │                      │
         └──────────┬───────────┘
                    ▼
              [Continues]
                    │
                    ▼
┌──────────────────┐
│  Task Complete   │
└──────────────────┘

2.5 Integration Points

  1. Slash Commands: 11 commands for workflow operations (v2.2.0+)
  2. CLAUDE.md Integration: Instructions loaded each session
  3. Configuration: .ab-method/structure/index.yaml
  4. Documentation Output: docs/architecture/ generated files
  5. Task Storage: tasks/[task-name]/ directories
  6. Agent System: .claude/agents/ specialized agents

3. Development Cycle Integration

3.1 Planning Phase

AB Method provides robust planning support:

Analysis Commands:

/analyze-project    # Full codebase analysis with parallel agents
/analyze-frontend   # Client-side deep-dive
/analyze-backend    # Server-side deep-dive

Task Definition with Enhanced Technical Context (v2.1.0+):

The /create-task command 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

Backend-First Planning: Mission ordering enforces backend work precedes frontend, establishing types and API contracts before UI implementation.

3.2 Execution Phase

Execution follows the sequential mission pattern with v2.3.0 flexibility:

Mission Commands:

/create-mission   # Transform tasks into focused missions
/resume-mission   # Continue incomplete missions
/extend-task      # Add missions to existing tasks (NEW in v2.3.0)

extend-task Use Case (v2.3.0):

# Scenario: Requirements changed after initial task creation
/extend-task

> "Which task to extend?"
> "user-authentication"

> "What new missions to add?"
> "Add OAuth integration with Google and GitHub"

# System adds new missions without disrupting completed work:
# ✓ Mission 1: Basic auth - COMPLETED
# ✓ Mission 2: Password reset - COMPLETED
# ○ Mission 3: OAuth Google - ADDED (extend-task)
# ○ Mission 4: OAuth GitHub - ADDED (extend-task)

Agent Deployment: Specialized agents auto-deploy based on mission type with intelligent selection (v2.1.0+):

  • Backend missions → nextjs-backend-architect
  • Frontend missions → shadcn-ui-adapter + UX expert
  • Testing missions → vitest-component-tester / playwright-e2e-tester

3.3 Testing Phase (BREAKING CHANGE in v2.0.0)

Dedicated Testing Workflow:

In v2.0.0, testing was separated into its own workflow instead of being part of implementation missions:

/test-mission     # Create comprehensive tests for a mission

Test Mission Flow:

Implementation Mission Complete
         │
         ▼
┌──────────────────┐
│ /test-mission    │
│                  │
│ 1. Load mission  │
│    outputs       │
│ 2. Deploy test   │
│    agents        │
│ 3. Generate      │
│    test suite    │
│ 4. Run tests     │
│ 5. Document      │
│    results       │
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│ Test Results     │
│ Documented       │
└──────────────────┘

3.4 Review Phase

Review integration through validation checkpoints:

Validation Pattern:

  • Tasks created with “Brainstormed” status
  • User validation required before implementation
  • Mission plans reviewed before execution

Documentation Updates:

/update-architecture    # Maintain architecture docs after changes

4. SOLID Principles Adherence

4.1 Single Responsibility Principle

Each component has focused responsibility: slash commands for invocation, core workflows for execution logic, agents for domain expertise, documentation for knowledge capture. The hierarchical Task→Mission structure enforces single-purpose work units.

The v2.0.0 separation of testing into a dedicated workflow further strengthens SRP by removing testing concerns from implementation missions.

Score: 4.2/5.0 (improved from 4.0 due to testing separation)

4.2 Open-Closed Principle

The agent system allows extension through new specialized agents without modifying core workflows. Configuration in index.yaml enables customization without code changes.

The v2.3.0 /extend-task command exemplifies OCP: existing tasks can be extended with new missions without modifying completed work.

However, adding new slash commands still requires file creation in the Claude commands directory.

Score: 3.8/5.0 (improved from 3.5 due to extend-task)

4.3 Liskov Substitution Principle

Missions are substitutable within tasks—any mission can be extended, resumed, or tested using the same command interface. Agents implement consistent invocation patterns.

The v2.2.0 individual slash commands maintain substitutability with the traditional /ab-master controller—both produce identical results.

Score: 3.7/5.0 (improved from 3.5 due to command parity)

4.4 Interface Segregation Principle

Slash commands are segregated by function: task commands, mission commands, analysis commands, and documentation commands. Users invoke only what they need.

The v2.2.0 individual commands represent ISP improvement: users no longer need the master controller for simple operations.

Score: 4.2/5.0 (improved from 4.0 due to individual commands)

4.5 Dependency Inversion Principle

The framework depends on Claude Code’s command system abstraction rather than specific implementation details. Agent dispatch uses configuration-driven selection.

The intelligent subagent integration (v2.1.0) improves DIP through dynamic agent selection based on mission characteristics rather than hardcoded assignments.

Score: 3.8/5.0 (improved from 3.5 due to intelligent agent selection)

4.6 Practical Examples

SRP Example - Testing Separation (v2.0.0):

# Before v2.0.0: Testing mixed with implementation
/create-mission  # Included tests

# After v2.0.0: Testing is separate concern
/create-mission  # Implementation only
/test-mission    # Testing only

OCP Example - Task Extension (v2.3.0):

# Extend task without modifying completed missions
/extend-task
> Add new OAuth missions to user-authentication task
# Completed missions unchanged, new missions appended

ISP Example - Individual Commands (v2.2.0):

# Before v2.2.0: Required master controller
/ab-master create-task

# After v2.2.0: Direct command access
/create-task

Overall SOLID Score: 3.9/5.0 (improved from 3.7)

Justification: AB Method v2.3.0 demonstrates improved SOLID adherence compared to v1.0.0. Single Responsibility is strengthened through the dedicated testing workflow separation in v2.0.0, ensuring implementation and testing concerns are properly decoupled. Interface Segregation improved significantly with v2.2.0’s individual slash commands, allowing users to access only the functionality they need without navigating through the master controller. Open-Closed principle adherence improved with v2.3.0’s extend-task capability, enabling task modification without altering completed work. The methodology nature of the framework means some SOLID principles apply less directly than in library code, but the design represents thoughtful evolution toward better architectural principles while maintaining the pragmatic user experience focus that made v1.0.0 successful. The incremental improvements across versions demonstrate commitment to architectural quality alongside feature development.


5. Production Readiness Assessment

5.1 Reliability

Maturity: Active development with 14 releases since August 2025. Single maintainer but consistent release cadence (semantic versioning with automated releases).

Error Handling: Mission resume capability (/resume-mission) enables recovery from incomplete work. Validation checkpoints prevent wasted effort. The v2.3.0 /extend-task allows scope changes without starting over.

Stability: Zero open issues suggests either limited adoption or stable implementation. CHANGELOG demonstrates systematic bug fixes (v2.1.3, v2.2.1, v2.2.2).

Score: 70/100 (improved from 65)

5.2 Observability

Agent Usage Tracking: Records which agents were deployed per mission with v2.1.0 enhanced tracking:

## 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

Progress Tracking: Task status lifecycle with clear transitions.

Documentation Generation: Auto-generated architecture files provide execution visibility.

Score: 75/100 (improved from 70)

5.3 Security

Minimal Attack Surface: File-based configuration with no external dependencies beyond Claude Code.

No Credential Handling: Framework does not manage authentication or secrets.

Local Execution: All operations execute locally within the Claude Code session.

Score: 75/100 (unchanged)

5.4 Performance

Token Efficiency: 60-70% token savings through sequential mission execution and context isolation.

Context Optimization: Each mission loads only necessary context (previous + current). Intelligent subagent integration (v2.1.0) improves agent selection efficiency.

Single-Session Design: Missions designed for single Claude session completion.

Score: 82/100 (improved from 80)

5.5 Maintainability

Documentation: README and CLAUDE.md provide comprehensive usage guidance. Architecture files auto-generated. CHANGELOG maintained with semantic-release automation.

Code Organization: Clear directory structure with separation between core, utils, agents, and structure.

Release Process: Automated semantic-release with GitHub Actions ensures consistent versioning.

Single Maintainer Risk: Limited bus factor with one primary contributor remains a concern.

Score: 70/100 (improved from 65)

Overall Production Score: 74/100 (improved from 71)

Dimension v1.0.0 Score v2.3.0 Score Change
Reliability 65/100 70/100 +5
Observability 70/100 75/100 +5
Security 75/100 75/100 -
Performance 80/100 82/100 +2
Maintainability 65/100 70/100 +5
Overall 71/100 74/100 +3

6. Best Practices & Patterns

AB Method codifies several development patterns:

Sequential Mission Pattern: One mission at a time with complete focus:

Mission 1: Complete entirely
     ↓
Mission 2: Build on Mission 1 knowledge
     ↓
Mission N: Accumulated context

Backend-First Full-Stack Pattern: Types flow from server to client:

Mission 1: Backend - Define todo model and API
Mission 2: Frontend - Use backend types in UI
Mission 3: Integration - Connect with type safety

Validation Checkpoint Pattern: Plans approved before implementation:

Task Created (Status: Brainstormed)
     ↓
User Reviews Plan
     ↓
User Approves (Status: Validated)
     ↓
Implementation Begins

Separated Testing Pattern (NEW in v2.0.0):

Implementation Mission Complete
     ↓
/test-mission
     ↓
Test Suite Generated
     ↓
Tests Executed
     ↓
Results Documented

Task Extension Pattern (NEW in v2.3.0):

Task with 3 Missions (2 Complete)
     ↓
/extend-task
     ↓
Requirements: Add OAuth support
     ↓
Task with 5 Missions (2 Complete, 3 New)
     ↓
Continue from Mission 3

Progressive Knowledge Building: Each mission enriches the context:

  • Mission outputs persist in tasks/[task-name]/
  • Later missions have access to earlier outputs
  • Architecture documentation accumulates

Agent-Domain Matching: Specialized agents deploy based on mission type, with intelligent selection (v2.1.0) considering mission characteristics.


7. Limitations & Trade-offs

Single Maintainer Risk: With only one contributor, long-term maintenance is uncertain. Bus factor of 1 represents significant project risk.

Technology Stack Coupling: Agents are optimized for Next.js, shadcn/ui, and SST. Teams using different stacks may find limited agent utility. No Django, Rails, or Vue agents available.

Sequential Execution Only: No support for parallel mission execution. Teams requiring concurrent work must manage outside the framework.

Test Coverage Unknown: No documented test coverage for the framework itself. Quality assurance relies on user testing and semantic-release CI.

Breaking Change History: The v2.0.0 testing separation was a breaking change, indicating the framework is still evolving its core patterns.

Documentation Gaps: While auto-generated architecture docs are a strength, user-facing documentation for advanced scenarios (multi-developer workflows, large projects) is limited.

Scalability Uncertainty: Performance with large projects (many tasks/missions) is not documented. Context accumulation may create pressure over extended projects. The extend-task feature (v2.3.0) may exacerbate this with very long-running tasks.

No Rollback Mechanism: If a mission introduces problems, there’s no built-in way to revert to a previous state beyond manual git operations.


8. Community & Ecosystem

AB Method has an emerging but growing community:

Adoption Metrics: Growing stars and forks indicate increasing adoption. Featured in awesome-claude-code listing provides discovery visibility.

Maintainer Activity: Active development with consistent release cadence (14 releases in 4 months). Creator (Ayoub Bensalah) publicly shares philosophy.

Issue Engagement: Zero open issues could indicate either stability or limited adoption depth. CHANGELOG shows responsive bug fixes.

Documentation: README provides comprehensive overview with examples. CHANGELOG maintained via semantic-release. No separate documentation site.

Related Projects: Part of the broader Claude Code extension ecosystem alongside BMAD Method, Simone, Claude Flow, and others. AB Method focuses specifically on mission-based incremental development.

npm Package: Available on npm (npx ab-method) for easy installation, improving discoverability and adoption.


9. Innovation & Differentiation

AB Method introduces several innovations distinguishing it from alternatives:

Mission-Based Decomposition: Unlike phase-based methodologies, AB Method decomposes work into discrete missions designed for single-session completion. This granularity enables precise context management and clear progress tracking.

Backend-First Philosophy: The explicit ordering of backend before frontend missions represents a novel approach to ensuring type safety in AI-assisted full-stack development. Types established in backend missions automatically inform frontend implementations.

60-70% Token Efficiency: Sequential mission execution with isolated context achieves substantial token savings compared to approaches that maintain full project context throughout development.

Pre-Installed Specialized Agents: 8 battle-tested agents for specific domains (shadcn/ui, Next.js, SST, testing) provide immediate domain expertise without configuration.

Dual Access Philosophy: Both direct slash commands (v2.2.0) and master controller invocation provide flexibility for different user preferences and workflow styles.

Separated Testing (v2.0.0): Breaking testing into its own workflow allows implementation and testing to be treated as distinct concerns with appropriate tooling.

Task Extension (v2.3.0): The ability to add missions to existing tasks without disrupting completed work addresses the common reality of evolving requirements.

Auto-Generated Architecture Documentation: Continuous documentation updates ensure project knowledge remains current, addressing a common pain point in AI-assisted development.

Comparison with Alternatives:

Feature AB Method BMAD Simone Claude Flow
Mission-based Yes Phase-based Rules-based Flow-based
Backend-first Yes Flexible N/A N/A
Specialized agents 8 included Custom N/A N/A
Individual commands 11 (v2.2.0) Personas Commands Commands
Task extension Yes (v2.3.0) N/A N/A N/A
Token savings 60-70% Varies Minimal Varies

10. References & Sources

  1. GitHub Repository: https://github.com/ayoubben18/ab-method
  2. npm Package: https://www.npmjs.com/package/ab-method
  3. CHANGELOG: https://github.com/ayoubben18/ab-method/blob/main/CHANGELOG.md
  4. awesome-claude-code: https://github.com/g-loot/awesome-claude-code - Community listing
  5. Next.js Documentation: https://nextjs.org/docs - Target framework
  6. shadcn/ui Documentation: https://ui.shadcn.com - Target UI library
  7. SST Documentation: https://sst.dev - Target infrastructure

11. Error Handling Patterns

11.1 Error Detection

Validation Checkpoints: User approval gates detect requirement misalignment before implementation waste.

Mission Status Tracking: Incomplete missions are detected through status lifecycle tracking:

Brainstormed → Validated → In Development → Testing → Completed

Interactive Clarification (v2.1.0): Intelligent subagent integration includes clarification prompts when mission requirements are ambiguous.

11.2 Error Recovery

Mission Resume: /resume-mission enables continuation from incomplete state:

# Resume incomplete mission
/resume-mission

> "Which mission to resume?"
> "todo-table/mission-2"

# Shows progress with agent tracking:
# ✓ Mission 1: Backend API - COMPLETED (nextjs-backend-architect)
# ⏳ Mission 2: Frontend Table - IN PROGRESS (shadcn-ui-adapter)
#   Last: Created base component with shadcn/ui patterns
#   Next: Add state management and data fetching

Task Resume: /resume-task recovers paused work at the task level.

Task Extension Recovery (v2.3.0): When requirements change mid-task, /extend-task allows adding missions without losing completed work:

/extend-task
> "Add OAuth missions to partially-completed auth task"
# Completed missions preserved, new missions appended

11.3 Error Reporting

Progress Documentation: Mission outputs document what was completed, enabling diagnosis of incomplete work.

Agent Usage Records: Records which agents were deployed, enabling review of execution approach.

Test Results Documentation (v2.0.0): /test-mission generates documented test results for post-mortem analysis.


12. Token Efficiency Analysis

12.1 Token Optimization Strategies

AB Method achieves 60-70% token savings through:

Context Isolation: Each mission loads only necessary context:

  • Previous mission outputs (accumulated knowledge)
  • Current mission specification
  • Relevant architecture documentation

Sequential Execution: No parallel context loading. Full context window dedicated to current mission.

Mission Scoping: Missions designed for single-session completion, preventing context accumulation within sessions.

Intelligent Agent Selection (v2.1.0): Agents load domain-specific knowledge relevant to their function, not full project state. Agent selection optimized based on mission characteristics.

12.2 Context Management

Progressive Knowledge Building:

Mission 1 Context: Project base + Mission 1 spec
Mission 2 Context: Project base + Mission 1 outputs + Mission 2 spec
Mission N Context: Project base + All previous outputs + Mission N spec

Documentation Persistence: Architecture docs persist to disk, available for selective loading rather than full context inclusion.

Agent Context: Specialized agents load domain-specific knowledge relevant to their function:

shadcn-ui-adapter: UI patterns + component library
nextjs-backend-architect: API patterns + backend conventions
vitest-component-tester: Testing patterns + coverage requirements

12.3 Token Efficiency Comparison

Approach Context Strategy Token Usage
Unstructured Full project always 100% (baseline)
Phase-based Phase context 50-70%
AB Method Mission context 30-40%

The 60-70% savings come from avoiding full project context loading on each interaction, instead loading only mission-relevant context plus accumulated outputs.


M. Methodology Principles

M.1 The Five Core Principles

AB Method is built on five coherent principles:

1. Singular Focus

  • One task at a time with laser focus
  • Missions execute sequentially, not in parallel
  • Context window dedicated to current work

2. Progressive Missions

  • Each mission builds on previous knowledge
  • Knowledge accumulates across boundaries
  • Eliminates redundant implementation

3. Backend-First

  • Types established in backend before frontend
  • API contracts precede UI work
  • Single source of truth for data models

4. Validation Checkpoints

  • Plans approved before implementation
  • Explicit approval gates
  • Prevents wasted effort

5. Ongoing Documentation

  • Architecture tracked continuously
  • Auto-generated documentation
  • Never becomes stale

M.2 Methodology Workflow

Complete lifecycle support through structured phases:

Phase Commands Activities
Analysis /analyze-* Codebase analysis, tech stack ID
Planning /create-task Task definition, mission planning
Execution /create-mission Sequential implementation
Testing /test-mission Unit and e2e testing (v2.0.0)
Extension /extend-task Add missions (v2.3.0)
Documentation /update-architecture Architecture updates

M.3 Task Lifecycle

Brainstormed → Validated → In Development → Testing → Completed
                              ↑
                              │
                    /extend-task (v2.3.0)
                    adds new missions

Each transition requires explicit action, ensuring deliberate progress through the development lifecycle. The v2.3.0 extend-task capability allows returning to “In Development” to add new missions without resetting completed work.

M.4 Evolution Summary (v1.0.0 → v2.3.0)

Principle v1.0.0 v2.3.0 Evolution
Focus Strong Strong Unchanged
Missions Sequential Sequential + Extensible Extended
Backend-First Yes Yes Unchanged
Validation Single gate Multi-gate Enhanced
Documentation Auto-gen Auto-gen + Agent tracking Enhanced
Testing Integrated Separated Breaking change

Version History

Version Date Key Changes
v1.0.0 Aug 2025 Initial release with 8 workflows
v1.1.0 Aug 2025 npm CLI installer
v2.0.0 Aug 2025 BREAKING: Testing separated
v2.1.0 Aug 2025 Intelligent subagent integration
v2.2.0 Sep 2025 Individual slash commands
v2.3.0 Nov 2025 extend-task command

Analysis completed following research-framework-analyzer skill methodology.


Generated: 2025-12-03 21:56 UTC | Template: framework-page.md.j2 v2.0 | Source Data

∵ RCR Regis ∴ | Claude Code Framework Research