A structured command system for AI-assisted Spec-Driven Development, enabling rapid feature delivery through systematic specifications, dependency-driven task execution via Beads, and brownfield/greenfield support.
/product:init- Initialize project with product documentation/product:roadmap- Define and prioritize features/product:progress- Track overall product development
/spec:create <feature>- Create unified spec (greenfield or brownfield with code detection)/spec:work [feature]- Implement next ready task using Beads and layer boundary testing/spec:status [feature]- Track feature progress via Beads
/change:propose <feature> <change>- Propose changes to existing features with delta tracking/change:complete <feature> <change>- Merge delta into spec and close Beads epic
/code:review [branch]- Comprehensive code review with architecture compliance/code:commit- Create smart git commits with conventional format/code:validate- Run build, lint, and test pipeline (multi-language)
# 1. Initialize project
/product:init my-project
# 2. Define features
/product:roadmap
# 3. Create & implement feature (unified workflow)
/spec:create authentication # Detects code, creates unified spec + Beads tasks
/spec:work authentication # Implements next ready task
/spec:work authentication # Continue until complete
/spec:status authentication # Check progress
# 4. Quality assurance
/code:validate
/code:review
/code:commit# 1. Create spec from existing code
/spec:create user-profile # Auto-detects existing code, documents it
# 2. Propose a change
/change:propose user-profile add-avatar # Creates delta.md with ADDED/MODIFIED/REMOVED
# 3. Implement the change
/spec:work user-profile # Works on change tasks from Beads
# 4. Complete the change
/change:complete user-profile add-avatar # Merges delta into spec.md
# 5. Quality assurance
/code:validate
/code:commit- Lightweight docs over heavy planning
- Dependency-driven task ordering via Beads
- AI-assisted implementation
- Prevention over debugging
- Greenfield & Brownfield support with code detection
- Layered Architecture: Router → Service → Repository → Entity
- Layer Boundary Testing: Test at component boundaries, not every method
- Stub→Test→Fix Pattern: Create stubs → Write tests → Implement
- Contextual Layer Detection: Only create tasks for layers actually needed
- Agent Specialization: Focused roles for complex tasks
docs/
├── product/ # Product vision & roadmap
├── spec/{feature}/ # Unified specs (requirements + design)
├── changes/ # Change proposals and deltas
│ └── {feature}/
│ └── {change}/
│ ├── proposal.md
│ └── delta.md
└── standards/ # Technical patterns & practices
commands/
├── product/ # Product management commands
├── spec/ # Feature development commands (unified workflow)
├── change/ # Change management commands
├── code/ # Code quality and git commands
└── README.md # Command documentation
agents/
├── analyst.md # Requirements gathering (code-aware)
├── architect.md # Technical design with Beads task generation
├── coder.md # Layer boundary testing implementation
├── scaffold.md # File scaffolding
├── context.md # Context retrieval
└── product.md # Progress tracking via Beads
.beads/ # Beads task tracking (git-backed)
└── beads.jsonl # Task and dependency data
- Beads Integration: Dependency-aware task tracking with
bdCLI - Code Detection: Auto-detects existing code in greenfield/brownfield modes
- Unified Specs: Single spec.md file with Requirements + Technical Design + Implementation Notes
- Change Management: Delta tracking with ADDED/MODIFIED/REMOVED structure
- Layer Boundary Testing: Test at component interfaces, not every method
- Contextual Tasks: Only create Beads tasks for layers actually needed
- Multi-Language: Auto-detects Node.js, Python, Rust, Go, Make projects
- Architecture Compliance: Automated pattern validation
- Smart Commits: Context-aware git messages
This system uses Beads for dependency-aware task management:
# Install Beads
npm install -g @beads/bd
# Initialize in project
bd init
# Common operations (handled by commands)
bd ready # Find next ready task
bd create "task" -l feat # Create task with labels
bd dep add <task> <blocks> --type blocks # Add dependency
bd close <task> # Mark completeBeads provides:
- Git-backed task storage (
.beads/beads.jsonl) - Dependency tracking (technical blocking relationships)
- Ready task detection (find unblocked work)
- JSON output for agent consumption
- Use unified workflow:
/spec:create→/spec:work→/spec:status - Leverage Beads: Let dependency tracking guide implementation order
- Detect existing code:
/spec:createworks for both greenfield and brownfield - Use change workflow:
/change:propose→/spec:work→/change:completefor modifications - Test at boundaries: Focus on layer interfaces, not internal methods
- Validate frequently: Use
/code:validatebefore commits - Track progress: Use
/spec:statusto monitor Beads tasks
This system transforms AI from requiring constant prompting into a team member that understands your standards, business context, implementation approach, and can handle both new features and changes to existing code.