9.4 KiB
🚀 BMAD Method - Complete Getting Started Guide
Master the enhanced BMAD workflow - From basic development to enterprise-grade quality engineering with collaborative AI agents.
🎯 Quick Start (2 Minutes)
Your First Story
# 1. Start development
*develop-story
# 2. Validate quality
*reality-audit
# 3. Auto-push (if Grade A) or get remediation options
That's it! The system handles the rest automatically.
🔧 Claude Code CLI Users - Premium Experience
Workspace Commands (All Agents)
*workspace-init # Initialize collaborative workspace
*workspace-status # Show active sessions & collaboration context
*workspace-handoff # Context-aware agent transitions
*workspace-cleanup # Automated maintenance & optimization
*workspace-sync # Restore context & synchronize state
Getting Started with Workspace
# 1. Initialize your project workspace
*workspace-init
# 2. Work on your story
*develop-story
# 3. Check workspace status anytime
*workspace-status
# 4. Hand off to QA with full context
*workspace-handoff qa
# 5. Clean up when done
*workspace-cleanup
What You Get:
- Automatic session lifecycle management
- Intelligent agent handoff suggestions
- Context preservation across sessions
- Built-in workspace health monitoring
- Enhanced productivity analytics
🤖 Choosing the Right Agent
🏗️ Development Phase Agents
James (Developer) - bmad dev
*develop-story # Systematic implementation
*reality-audit # Quality validation
*build-context # Pre-fix investigation
*workspace-handoff # Context-aware transitions
Use for: Code implementation, debugging, technical problem-solving
Quinn (QA) - bmad qa
*reality-audit # Manual quality audit
*audit-validation # Auto-remediation with fix stories
*story-code-audit # Cross-reference stories vs code
*workspace-cleanup # Quality-focused maintenance
Use for: Quality validation, testing, issue remediation
Morgan (Scrum Master) - bmad sm
*create-story # Generate development stories
*reality-audit # Story quality validation
*workspace-status # Team collaboration overview
Use for: Story creation, sprint planning, team coordination
🎨 Planning Phase Agents
Alex (Analyst) - bmad analyst
*create-research # Deep domain investigation
*workspace-init # Research session tracking
Use for: Market research, user analysis, requirement gathering
Sam (Architect) - bmad architect
*create-architecture # Technical design documents
*workspace-handoff # Design context transfer
Use for: System design, technology decisions, architecture planning
Jordan (UX Expert) - bmad ux-expert
*create-ux # UI/UX specifications
*workspace-sync # Design context restoration
Use for: User experience design, interface planning, usability
📊 Product Management Agents
John (Product Manager) - bmad pm
*create-prd # Product Requirements Documents
*workspace-status # Product management overview
Use for: PRD creation, product strategy, feature prioritization
Sarah (Product Owner) - bmad po
*validate-story-draft # Story validation & acceptance criteria
*workspace-cleanup # Backlog maintenance
Use for: Backlog management, story refinement, acceptance criteria
🎯 Quality System Mastery
Understanding Your Quality Score
Grade A (90-100) - Exceptional ✅
- Auto-push eligible
- Zero remediation needed
- Production ready
Grade B (80-89) - Good 🔵
- Meets quality gates
- Minor improvements suggested
- Ready for deployment
Grade C (70-79) - Acceptable 🟡
- Needs improvement before production
- Specific remediation options provided
- Safe to continue development
Grade D (60-69) - Poor 🟠
- Remediation required
- Multiple issues detected
- Fix before proceeding
Grade F (<60) - Failing 🔴
- Major issues detected
- Comprehensive remediation needed
- Do not deploy
Automatic Features That Save Time
🔄 Loop Detection (After 3 Failed Attempts)
# System automatically provides:
"Copy this prompt to Gemini/GPT-4/Claude for collaboration:
[Generated external collaboration prompt]"
🤖 Automatic Remediation (Zero Commands)
- Quality issues → Fix stories generated automatically
- Oversized stories → Auto-split into manageable pieces
- Mixed concerns → Surgical fixes created immediately
📤 Intelligent Git Push (Grade A Only)
- Automatic push with quality metrics in commit message
- Manual override available with
*Push2Git
🎛️ Role-Optimized Performance
Agent Temperature Settings (Automatic)
Development Agents (Precise Code)
- Developer: 0.4 - Focused, accurate implementations
- QA: 0.3 - Systematic, detailed analysis
Creative Agents (Innovation)
- Analyst: 0.8 - High creativity for market insights
- UX Expert: 0.75 - Creative interface solutions
Strategic Agents (Balanced)
- Architect: 0.6 - Structured creativity for design
- PM: 0.7 - Strategic thinking with clear communication
Management Agents (Organized)
- Scrum Master: 0.5 - Balanced planning and execution
- Product Owner: 0.6 - User-focused with systematic approach
You don't need to configure these - they're optimized automatically per agent role.
🔧 IDE Integration Features
Automatic IDE Detection
The system automatically detects and optimizes for:
- Claude Code - Premium workspace experience
- Cursor - Native git panels, integrated testing
- Windsurf - Seamless tool integration
- Cline, Trae, Roo - Enhanced command batching
- GitHub Copilot - Compatible workflow integration
What Changes Per IDE
Claude Code CLI Users:
- Get native workspace commands
- Automatic session management
- Context-aware handoffs
- Built-in maintenance system
Other IDEs:
- Optimized tool usage patterns
- Reduced approval prompts
- Intelligent command batching
📋 Advanced Workflows
Story-to-Code Auditing
# After completing stories, validate implementation
*story-code-audit
# System automatically:
# 1. Cross-references completed story tasks vs actual code
# 2. Identifies gaps between planned vs implemented features
# 3. Generates remediation stories for missing functionality
Regression Prevention
# Before making changes to existing code
*build-context
# System provides:
# - Git history analysis
# - Risk assessment
# - Pattern compliance validation
# - Safe modification strategies
External Collaboration (Automatic)
# When stuck after 3 attempts, system provides:
"Escalation: Copy this prompt to external AI:
CONTEXT: [Current situation]
ATTEMPTED SOLUTIONS: [What's been tried]
SPECIFIC CHALLENGE: [Exact issue]
DESIRED OUTCOME: [What success looks like]
Please provide alternative approaches or solutions."
⚡ Pro Tips for Maximum Efficiency
1. Start with Workspace (Claude Code CLI)
*workspace-init # Sets up collaborative environment
2. Use the Right Agent for the Phase
- Planning: analyst → pm → architect → ux-expert
- Development: sm → dev → qa
- Management: po for backlog, sm for sprints
3. Trust the Automation
- Let automatic remediation handle quality issues
- Use loop detection for external collaboration
- Rely on Grade A auto-push for perfect completions
4. Leverage Context Handoffs
*workspace-handoff architect # Intelligent context transfer
5. Monitor Quality Continuously
*reality-audit # Run anytime for quality insights
🚨 Troubleshooting Common Issues
"My story is too large"
Solution: System automatically detects and splits oversized stories
- Grade D/F audits trigger automatic story splitting
- Each piece becomes manageable (<8 tasks)
"I'm stuck in a debugging loop"
Solution: Loop detection activates after 3 failed attempts
- Provides external AI collaboration prompts
- Resets counter on successful progress
"Quality score is low"
Solution: Automatic remediation executes immediately
- Fix stories generated without manual commands
- Specific improvement actions provided based on grade
"Agent handoff lost context"
Solution: Use workspace commands (Claude Code CLI)
*workspace-handoff [agent] # Preserves full context
"Don't know what to do next"
Solution: System provides automatic options based on audit grade
- Grade-specific actions with effort estimates
- Clear next steps eliminate confusion
📈 Success Metrics to Track
Quality Indicators
- Grade A stories: Target 70%+ for mature projects
- Automatic remediation: Should handle 80%+ of quality issues
- Loop detection: Should rarely activate (<5% of stories)
Productivity Indicators
- Story completion time: 60+ minutes saved per debugging session
- Manual commands: Zero needed for remediation execution
- Context handoffs: Seamless transitions between agents
Process Indicators
- Build status: Clean builds with zero warnings
- Regression issues: <10% of stories cause regressions
- Simulation patterns: <5% reach production
🎯 Ready to master AI-assisted development? Follow these patterns and let the enhanced BMAD framework handle the complexity while you focus on creating great software.