bmad-skill

BMAD-CORE (Collaboration Optimized Reflection Engine) is a revolutionary framework that amplifies human potential through specialized AI agents. Unlike traditional AI tools that replace human thinking, BMAD-CORE guides through reflective workflows that bring out the best ideas in both human and AI.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "bmad-skill" with this command: npx skills add zpankz/mcp-skillset/zpankz-mcp-skillset-bmad-skill

BMAD v6-Alpha Skill

What is BMAD-CORE?

BMAD-CORE (Collaboration Optimized Reflection Engine) is a revolutionary framework that amplifies human potential through specialized AI agents. Unlike traditional AI tools that replace human thinking, BMAD-CORE guides through reflective workflows that bring out the best ideas in both human and AI.

Core Philosophy:

  • Collaboration: Human-AI partnership where both contribute unique strengths

  • Optimized: Refined processes for maximum effectiveness

  • Reflection: Guided thinking that unlocks better solutions

  • Engine: Powerful framework orchestrating specialized agents and workflows

BMAD v6-Alpha Architecture

Module System

BMAD organizes capabilities into four main modules:

  1. Core Module (Always Installed)

Foundation powering all modules:

  • BMad Master Agent - Universal orchestrator and task executor

  • Party Mode - Multi-agent collaborative sessions

  • Brainstorming Workflow - Core ideation capabilities

  • Configuration System - User preferences and settings

Location: bmad/core/

  1. BMM Module - BMad Method (Software Development)

Revolutionary agile software development with Scale Adaptive Workflow Engine™:

  • 10 Specialized Agents: Analyst, PM, Architect, SM, Dev, Game Designer, Game Architect, Game Dev, UX Designer, TEA (Test Architect)

  • Four-Phase Methodology: Analysis → Planning → Solutioning → Implementation

  • Scale Levels 0-4: From single changes to enterprise projects

  • Just-In-Time Design: Create specs one epic at a time

  • Story State Machine: Automated story progression (BACKLOG → TODO → IN PROGRESS → DONE)

Location: bmad/bmm/ (if installed)

  1. BMB Module - BMad Builder (Custom Development)

Tools for creating custom agents, workflows, and modules:

  • Builder Agent - Guides creation of custom components

  • 7 Builder Workflows: create-agent, create-workflow, create-module, audit-workflow, edit-workflow, module-brief, redoc

  • Three Agent Types: Full module agents, hybrid agents, standalone agents

Location: bmad/bmb/ (if installed)

  1. CIS Module - Creative Intelligence Suite

Innovation and creative problem-solving:

  • 5 Specialized Agents: Brainstorming Coach (Carson), Design Thinking Coach (Maya), Problem Solver (Dr. Quinn), Innovation Strategist (Victor), Storyteller (Sophia)

  • 150+ Creative Techniques: Across brainstorming, design thinking, problem-solving, innovation strategy, storytelling

  • Energy-Aware Sessions: Adapts to user engagement

Location: bmad/cis/ (shared resources appear even if not explicitly selected)

  1. BMD Module - BMad Development (Project Management)

Internal tools for BMAD development itself:

  • CLI Chief - Command-line management

  • Doc Keeper - Documentation management

  • Release Chief - Release management

Location: bmd/ (development only)

Installation Structure

All modules install to a single unified bmad/ folder:

your-project/ └── bmad/ ├── core/ # Core framework (always present) ├── bmm/ # BMad Method (if selected) ├── bmb/ # BMad Builder (if selected) ├── cis/ # Creative Intelligence (shared resources) ├── _cfg/ # Your customizations │ ├── agents/ # Agent sidecar files │ └── manifest.yaml └── [module]/config.yaml # Per-module configuration

Working with BMAD Agents

Agent Activation Methods

Method 1: Claude Code Slash Commands (Recommended)

When BMAD is installed locally in .claude/commands/bmad/ , use these slash commands:

Core Module

/bmad:bmad:core:agents:bmad-master /bmad:bmad:core:workflows:brainstorming /bmad:bmad:core:workflows:party-mode

BMM Module Agents

/bmad:bmad:bmm:agents:analyst # Business Analyst - Phase 1 /bmad:bmad:bmm:agents:pm # Product Manager - Phase 2 /bmad:bmad:bmm:agents:architect # Architect - Phase 3 /bmad:bmad:bmm:agents:sm # Scrum Master - Phase 4 /bmad:bmad:bmm:agents:dev # Developer - Phase 4 /bmad:bmad:bmm:agents:game-designer # Game projects /bmad:bmad:bmm:agents:game-architect # Game architecture /bmad:bmad:bmm:agents:game-dev # Game development /bmad:bmad:bmm:agents:ux-designer # UX design /bmad:bmad:bmm:agents:tea # Test Architect

BMB Module

/bmad:bmad:bmb:agents:bmad-builder # Builder agent

BMB Workflows (Direct Access)

/bmad:bmad:bmb:workflows:create-agent /bmad:bmad:bmb:workflows:create-workflow /bmad:bmad:bmb:workflows:create-module /bmad:bmad:bmb:workflows:audit-workflow /bmad:bmad:bmb:workflows:edit-workflow /bmad:bmad:bmb:workflows:module-brief /bmad:bmad:bmb:workflows:redoc /bmad:bmad:bmb:workflows:convert-legacy

BMD Module (Development/Management)

/bmad:bmad:bmd:agents:cli-chief /bmad:bmad:bmd:agents:doc-keeper /bmad:bmad:bmd:agents:release-chief

Local Installation Location: /Users/mikhail/.claude/commands/bmad/

Method 2: Direct File Loading

  • Drag and drop agent file from .claude/commands/bmad/

  • Use @ mention with agent file path

  • Navigate to agent file in your IDE's file browser

Method 3: BMad Master Orchestrator Load BMad Master first, then use menu to access specific agents or workflows.

Agent Structure

All BMAD agents follow a consistent YAML structure:

agent: metadata: id: bmad/[module]/agents/[name].md name: AgentName title: Agent Title icon: 🎯 module: [bmm|bmb|cis|core]

persona: role: Primary Role + Secondary Expertise identity: Background and experience description communication_style: How the agent communicates principles: - Guiding principle 1 - Guiding principle 2

menu: - trigger: command-name workflow: "{project-root}/bmad/[module]/workflows/[workflow]/workflow.yaml" description: What this command does

Common Agent Commands

Every agent has these built-in commands:

  • *help

  • Show agent's menu

  • *exit

  • Exit agent with confirmation

Module-specific commands use asterisk prefix:

  • Analyst: *workflow-status , *brainstorm-project , *research , *product-brief

  • PM: *workflow-status , *prd , *tech-spec , *gdd (games)

  • Architect: *architecture , *tech-spec

  • SM: *create-story , *story-ready , *story-context , *story-done , *retrospective

  • Dev: *dev-story , *review-story , *story-done

BMM: The Four-Phase Software Development Methodology

Universal Entry Point: workflow-status

ALWAYS START HERE! Before beginning any workflow:

Load Analyst or PM agent, then:

*workflow-status

What it does:

  • ✅ Checks for existing workflow status file

  • ✅ Displays current phase, progress, and next action

  • ✅ Guides new projects to appropriate workflows

  • ✅ Routes brownfield projects to documentation first

  • ✅ Provides clear recommendations for next steps

Phase 1: Analysis (Optional)

Purpose: Project discovery and requirements gathering

Agents: Analyst, Game Designer

Workflows:

  • brainstorm-project

  • Software solution exploration

  • brainstorm-game

  • Game concept ideation (5 methodologies)

  • research

  • Multi-mode research (market/technical/deep)

  • product-brief

  • Strategic product planning

  • game-brief

  • Structured game design foundation

When to use: New projects needing concept development

Phase 2: Planning (Required)

Purpose: Scale-adaptive planning with appropriate documentation

Agents: PM, Game Designer

Scale Levels:

Level Scope Stories Outputs Next Phase

0 Single atomic change 1 tech-spec.md + story Implementation

1 Simple feature 2-10 tech-spec.md + epic + stories Implementation

2 Focused project 5-15 (1-2 epics) PRD.md + epics.md Tech-spec → Implementation

3 Complex project 12-40 (2-5 epics) PRD.md + epics.md Solutioning → Implementation

4 Enterprise scale 40+ (5+ epics) PRD.md + epics.md Solutioning → Implementation

Software Projects:

  • Level 0-1: *tech-spec (Architect) → Implementation

  • Level 2: *prd (PM) → *tech-spec → Implementation

  • Level 3-4: *prd (PM) → Solutioning → Implementation

Game Projects:

  • All Levels: *gdd (Game Designer) → Optional narrative → Solutioning or Implementation

Key Outputs:

  • bmm-workflow-status.md

  • Versioned workflow state with story backlog

  • PRD.md

  • Product Requirements (L2-4)

  • epics.md

  • Epic breakdown (L2-4)

  • tech-spec.md

  • Technical specification (L0-2)

  • story-*.md

  • User story files

Phase 3: Solutioning (Levels 3-4 Only)

Purpose: Architecture and technical design for complex projects

Agents: Architect, Game Architect

Workflows:

  • architecture

  • Create overall architecture.md with ADRs

  • tech-spec

  • Create epic-specific tech specs (JIT - Just-In-Time)

Just-In-Time Approach:

FOR each epic in sequence: WHEN ready to implement epic: Architect: Run *tech-spec for THIS epic only → Creates tech-spec-epic-N.md IMPLEMENT epic completely THEN move to next epic

Critical: Tech specs created ONE AT A TIME, not all upfront. This prevents over-engineering and incorporates learning.

Phase 4: Implementation (Iterative)

Purpose: Transform requirements into working software

Agents: SM (Scrum Master), Dev (Developer)

The Story State Machine:

BACKLOG → TODO → IN PROGRESS → DONE

State Definitions:

  • BACKLOG: Ordered list of stories awaiting drafting (populated at phase transition)

  • TODO: Single story needing drafting or awaiting approval

  • IN PROGRESS: Single story approved for development

  • DONE: Completed stories with dates and points

Implementation Loop:

  1. SM: *create-story (drafts story from TODO) → Story file created with Status="Draft"

  2. USER REVIEWS STORY

  3. SM: *story-ready (approves story) → TODO → IN PROGRESS → BACKLOG → TODO (next story) → Story Status = "Ready"

  4. SM: *story-context (optional but recommended) → Generates expertise injection XML

  5. DEV: *dev-story (implements story) → Reads IN PROGRESS section → Implements with context guidance

  6. USER REVIEWS IMPLEMENTATION

  7. DEV: *story-done (marks complete) → IN PROGRESS → DONE → TODO → IN PROGRESS (if exists) → BACKLOG → TODO (if exists) → Story Status = "Done"

  8. REPEAT until all stories complete

  9. SM: *retrospective (after epic complete) → Captures learnings for next epic

Key Workflows:

  • create-story

  • SM drafts story from TODO

  • story-ready

  • SM approves story for dev (after user review)

  • story-context

  • SM generates contextual expertise

  • dev-story

  • DEV implements story

  • story-done

  • DEV marks story complete (after user confirms DoD)

  • review-story

  • Quality validation (optional)

  • correct-course

  • Handle issues/changes

  • retrospective

  • Capture epic learnings

Critical Rule: Agents NEVER search for "next story" - they ALWAYS read from status file.

Working with Workflows

Workflow Execution

Workflows are executed through agent menus:

  • Load agent (Analyst, PM, Architect, SM, Dev, etc.)

  • Agent shows menu with available commands

  • Enter command (e.g., *prd , *create-story )

  • Agent loads and executes workflow.yaml

  • Workflow guides through structured steps

Workflow Structure

All workflows are YAML-based with consistent structure:

workflow: metadata: id: workflow-unique-id name: Workflow Name version: 1.0.0

config: output_folder: "{output_folder}"

steps: - name: Step Name prompt: | Detailed instructions for this step

Key Workflow Patterns

Validation Workflows: Use validate-workflow handler

  • trigger: validate-tech-spec validate-workflow: "{project-root}/bmad/bmm/workflows/2-plan-workflows/tech-spec/workflow.yaml" checklist: "{project-root}/bmad/bmm/workflows/2-plan-workflows/tech-spec/checklist.md" document: "{output_folder}/tech-spec.md"

Data-Driven Workflows: Use data parameter

  • trigger: command-name workflow: "{project-root}/bmad/[module]/workflows/[name]/workflow.yaml" data: "{output_folder}/context.md"

BMB: Building Custom Components

Creating Custom Agents

Load BMad Builder agent, then use:

  • *create-agent

  • Design and implement custom agents

  • *create-workflow

  • Build new workflow definitions

  • *create-module

  • Design custom modules

Agent Types

  1. Full Module Agents: Complete agents embedded in modules
  • Location: src/modules/[module]/agents/

  • Example: PM, Analyst, Architect

  1. Hybrid Agents: Shared across modules
  • Location: Shared agent directories

  • Example: Brainstorming agents used by multiple modules

  1. Standalone Agents: Tiny, specialized agents
  • Location: Independent of modules

  • Example: Single-purpose utility agents

Builder Workflows

  • audit-workflow

  • Review and validate workflows

  • edit-workflow

  • Modify existing workflows

  • module-brief

  • Document module specifications

  • redoc

  • Regenerate documentation

  • convert-legacy

  • Migrate legacy formats

CIS: Creative Intelligence

CIS Agents and Workflows

5 Specialized Agents with unique personas:

  • Carson - Brainstorming Coach (energetic facilitator)

  • Maya - Design Thinking Maestro (jazz-like improviser)

  • Dr. Quinn - Problem Solver (detective-scientist)

  • Victor - Innovation Strategist (strategic precision)

  • Sophia - Storyteller (whimsical narrator)

5 Interactive Workflows:

  • brainstorming

  • 36 creative techniques across 7 categories

  • design-thinking

  • 5-phase human-centered design

  • problem-solving

  • Root cause analysis and solution generation

  • innovation-strategy

  • Business model innovation

  • storytelling

  • 25 story frameworks

Using CIS

CIS workflows can be:

  • Invoked directly via CIS agents

  • Embedded in other modules (BMM uses brainstorming)

  • Used standalone for creative sessions

Best Practices

DO's

✅ Always start with *workflow-status

  • Universal entry point ✅ Respect the scale levels - Don't over-document small projects ✅ Use Just-In-Time design - Create tech specs one epic at a time ✅ Follow the story state machine - Never skip state transitions ✅ Run retrospectives - After each epic for continuous learning ✅ Generate story-context - Provides targeted expertise ✅ Validate before advancing - User reviews stories and implementations ✅ Document brownfield first - Never plan without understanding existing code

DON'Ts

❌ Don't skip workflow-status check - It guides your entire journey ❌ Don't create all tech specs upfront - Use JIT approach ❌ Don't batch story creation - One at a time through state machine ❌ Don't manually track stories - Let state machine manage progression ❌ Don't plan brownfield without docs - Run analysis first ❌ Don't over-document Level 0-1 - Simple projects need minimal docs ❌ Don't skip Phase 3 for Level 3-4 - Architecture is critical

Configuration and Customization

Module Configuration

Each module has config.yaml with:

project_name: "Your Project" output_folder: "./docs/bmad" user_name: "Your Name" communication_language: "English" technical_level: "intermediate" # BMM specific

Location: bmad/[module]/config.yaml

Agent Customization (Sidecar Files)

Customize any agent via sidecar files in bmad/_cfg/agents/ :

bmad/_cfg/agents/core-bmad-master.customize.yaml

customizations: persona: name: "CustomName" communication_style: "Your preferred style"

menu: - trigger: custom-command workflow: "{project-root}/path/to/custom/workflow.yaml" description: "Custom functionality"

Update-Safe: Customizations persist through BMAD updates

Multi-Language Support

Agents communicate in configured language:

  • Set in module config.yaml : communication_language: "Spanish"

  • Agents automatically adapt communication

  • Documentation output can use different language

Troubleshooting Common Issues

"Workflow status file not found"

→ Run *workflow-status first to initialize

"Story not in expected state"

→ Check bmm-workflow-status.md

  • verify story is in correct section

"Can't find next story"

→ Agents don't search - they read from status file sections

"Brownfield planning fails"

→ Run documentation workflow first (coming soon: brownfield-analysis)

"Over-documented small project"

→ Check scale level - Level 0-1 should skip PRD and architecture

"Tech specs created all at once"

→ Use JIT approach - one epic at a time during implementation

Quick Reference

Essential Commands by Agent

Analyst:

  • *workflow-status

  • Check status

  • *brainstorm-project

  • Ideation

  • *research

  • Market/technical research

  • *product-brief

  • Product strategy

PM:

  • *workflow-status

  • Check status

  • *prd

  • Create PRD (L2-4)

  • *tech-spec

  • Create tech spec (L0-2)

  • *gdd

  • Game design document

Architect:

  • *architecture

  • Overall architecture (L3-4)

  • *tech-spec

  • Epic-specific specs (JIT)

SM:

  • *create-story

  • Draft story

  • *story-ready

  • Approve story

  • *story-context

  • Generate expertise

  • *retrospective

  • Capture learnings

Dev:

  • *dev-story

  • Implement story

  • *review-story

  • Quality check

  • *story-done

  • Mark complete

BMad Builder:

  • *create-agent

  • Build custom agent

  • *create-workflow

  • Build workflow

  • *create-module

  • Build module

File Locations

Config: bmad/[module]/config.yaml

Agents: bmad/[module]/agents/

Workflows: bmad/[module]/workflows/

Tasks: bmad/core/tasks/

Customizations: bmad/_cfg/agents/

Status Tracking: {output_folder}/bmm-workflow-status.md

Support Resources

  • 💬 Discord Community

  • 🐛 Issue Tracker

  • 🎥 YouTube Tutorials

  • 📚 Full Documentation

Framework: BMAD-CORE v6-alpha | Version: Alpha | Branch: v6-alpha

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Automation

agent-observability

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

subagent-prompt-construction

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

agent prompt evolution

No summary provided by upstream source.

Repository SourceNeeds Review
Automation

mcp_agent_mail

No summary provided by upstream source.

Repository SourceNeeds Review