Structured AI-assisted software development — from idea to shipped code.
BMAD (Business Method for Agile Development) is a structured AI-first product development framework. This plugin provides a full drop-in Agent Zero integration: 20 specialized agent personas, 5 global skills, and the complete workflow library.
1. Install the plugin
git clone https://github.com/vanja-emichi/bmad_method.git usr/plugins/bmad-method2. Initialize a BMAD project
Select the BMad Master profile in Agent Zero, then run:
bmad init
This creates the project workspace and initializes project-scoped memory.
3. Start building
Tell BMad Master what you want to build. It routes you to the right specialist for your current phase — no manual agent selection needed.
BMAD organizes development into phases — Ideation → Planning → Architecture → Implementation → Testing. Each phase has specialist agents who own it.
You talk to BMad Master, the orchestrator. It reads project state and routes you to the right specialist. Specialists produce artifacts that the next phase consumes. State persists across sessions.
User → BMad Master → [routes to specialist] → artifact produced → state updated → next phase
| Module | Skill | Purpose |
|---|---|---|
| BMM — Business Method Module | bmad-bmm |
Full product lifecycle: discovery → planning → architecture → implementation |
| BMB — Builder Module | bmad-bmb |
Meta-module for creating and extending BMAD agents, workflows, and modules |
| TEA — Testing Excellence Accelerator | bmad-tea |
Test architecture, ATDD, automation, CI, NFR assessment |
| CIS — Creative Intelligence Suite | bmad-cis |
Innovation strategy, design thinking, storytelling, structured problem solving |
| Agent | Persona | Commands | Module |
|---|---|---|---|
| bmad-master | BMad Master 🧙 | LT LW AE DG PM |
Core |
| bmad-analyst | Mary 📊 | BP MR DR TR CB WB DP GPC |
BMM |
| bmad-pm | John 📋 | CP VP EP CE IR CC |
BMM |
| bmad-ux-designer | Sally 🎨 | CU |
BMM |
| bmad-architect | Winston 🏗️ | CA IR |
BMM |
| bmad-dev | Amelia 💻 | DS CR CK |
BMM |
| bmad-qa | Quinn 🧪 | QA |
BMM |
| bmad-sm | Bob 🏃 | SP CS VS SS ER CC |
BMM |
| bmad-quick-dev | Barry 🚀 | QS QQ CR |
BMM |
| bmad-tech-writer | Paige 📚 | DP WD US MG VD EC |
BMM |
| bmad-test-architect | Murat 🔬 | TMT TF AT TA TD TR NR CI RV |
TEA |
| bmad-brainstorming-coach | Carson 🧠 | BS |
CIS |
| bmad-design-thinking | Maya 🎭 | DT |
CIS |
| bmad-innovation | Victor ⚡ | IS |
CIS |
| bmad-storyteller | Sophia 📖 | ST |
CIS |
| bmad-problem-solver | Dr. Quinn 🔬 | PS |
CIS |
| bmad-presentation | Caravaggio 🎨 | SD EX PD CT IN VM CV |
CIS |
| bmad-agent-builder | Bond 🤖 | BA EA VA VS VF |
BMB |
| bmad-workflow-builder | Wendy 🔄 | BW EW VW MV RW CW |
BMB |
| bmad-module-builder | Morgan 🏗️ | PB CM EM VM IM |
BMB |
Each BMAD module ships as an Agent Zero skill containing a SKILL.md manifest, a module-help.csv routing table, and a workflows/ directory with step files. Skills are loaded on-demand — agents call skills_tool:load to get workflow instructions, then execute them exactly as defined. No workflow logic lives in agent prompts.
Each agent uses A0's native prompt slot system for clean separation of concerns:
┌─────────────────────────────────────────────────────────┐
│ role.md + tips.md (WHO — persona, principles) │
├─────────────────────────────────────────────────────────┤
│ specifics.md (HOW — BMAD framework rules) │
├─────────────────────────────────────────────────────────┤
│ solving.md (WHEN — workflow execution pattern)│
├─────────────────────────────────────────────────────────┤
│ communication.md (FORMAT — JSON output rules) │
├─────────────────────────────────────────────────────────┤
│ SKILL.md (WHAT — workflow routing + steps) │
├─────────────────────────────────────────────────────────┤
│ .a0proj/instructions/ (WHERE — project config + state) │
└─────────────────────────────────────────────────────────┘
role.md defines persona and principles. specifics.md carries the BMAD operational framework (moved from communication.md to use A0's dedicated slot). solving.md overrides the default problem-solving pattern with BMAD's workflow-first execution model. communication.md handles JSON format rules only. SKILL.md contains workflow routing. Project instructions provide config and state.
| Extension | Hook | Purpose |
|---|---|---|
_80_bmad_routing_manifest.py |
message_loop_prompts_after |
Dynamically builds routing manifest from skill routing tables with artifact detection and staleness warnings |
- Project-level FAISS store — single vector database shared by all agents (structural isolation)
- Knowledge preload — recursive FAISS scan on agent init
- No per-agent stores — cross-agent recall by design
BMAD ships a live project status dashboard. After installation, the BMAD button appears in Agent Zero's sidebar. The dashboard is read-only — it observes agent state without writing to it.
| Document | Description |
|---|---|
| Quick Start | Clone and initialize a BMAD project |
| Document Lifecycle | Artifact staleness detection and consistency checks |
| CHANGELOG | Full version history |
| BMAD-METHOD Core | Upstream framework architecture reference |
| Agent Zero | Host platform |
Plugin: 1.0.8 | Upstream: Core 6.2.2 · BMB 1.5.0 · TEA 1.9.1 · CIS 0.1.9
Upstream parity: ~98% — full coverage across BMM, BMB, TEA, CIS modules. Only BMGD (game dev) module deferred. Verified against bmad-code-org/BMAD-METHOD.
See CHANGELOG.md for full version history.
- Agent Zero (latest stable release)
- An LLM with large context window recommended (Claude Sonnet or better)
- BMAD-METHOD Core
- bmad-builder (BMB)
- bmad-method-test-architecture-enterprise (TEA)
- bmad-module-creative-intelligence-suite (CIS)
MIT
