Commands Overview

Complete reference for all SDD commands in Agent Teams Lite

Agent Teams Lite provides 9 slash commands for managing the Spec-Driven Development workflow. Each command delegates work to specialized sub-agents that operate with fresh context.

Available Commands

/sdd-init

Initialize SDD in your project

/sdd-explore

Investigate ideas and approaches

/sdd-new

Start a new change

/sdd-continue

Run next phase

/sdd-ff

Fast-forward planning

/sdd-apply

Implement tasks

/sdd-verify

Validate implementation

/sdd-archive

Close and archive change

Command Categories

Initialization

• /sdd-init - Bootstrap SDD in your project. Run this once per project to detect stack and setup persistence.

Planning

• /sdd-explore - Research and investigate without creating files. Compare approaches, identify risks.
• /sdd-new - Start a new change. Runs exploration then creates proposal.
• /sdd-ff - Fast-forward through all planning phases (proposal → specs → design → tasks).
• /sdd-continue - Run the next phase in the dependency chain.

Implementation

• /sdd-apply - Write code following specs and design. Supports TDD workflow.

Validation

• /sdd-verify - Validate implementation against specs. Runs tests and produces compliance matrix.

Completion

• /sdd-archive - Merge specs and archive the completed change.

Typical Workflows

Full Structured Workflow

/sdd-init                  # One-time setup
/sdd-new add-dark-mode     # Start change
# Review proposal
/sdd-continue              # Generate specs
/sdd-continue              # Generate design
/sdd-continue              # Generate tasks
/sdd-apply                 # Implement
/sdd-verify                # Validate
/sdd-archive               # Close

Fast Planning Workflow

/sdd-init
/sdd-ff add-dark-mode      # Create proposal, specs, design, tasks
/sdd-apply                 # Implement
/sdd-verify                # Validate
/sdd-archive               # Close

Exploration Only

/sdd-init
/sdd-explore "state management options"  # Research without creating change
# Review analysis, then decide
/sdd-new implement-zustand                # Start change if approved

Command Characteristics

Command	Creates Artifacts	Modifies Code	User Approval
/sdd-init	Yes (config)	No	No
/sdd-explore	No	No	No
/sdd-new	Yes (proposal)	No	Between phases
/sdd-continue	Yes (next artifact)	No	Between phases
/sdd-ff	Yes (4 artifacts)	No	After all phases
/sdd-apply	Yes (progress)	Yes	Between task batches
/sdd-verify	Yes (report)	No	Before archive
/sdd-archive	Yes (archive)	No	No

Sub-Agent Delegation

All SDD commands work by delegating to specialized sub-agents via the Task tool. This architecture keeps the orchestrator lightweight and gives each phase fresh context.

YOU → /sdd-new add-csv-export
      ↓
  ORCHESTRATOR (delegate-only)
      ↓
      ├→ EXPLORER sub-agent    → returns analysis
      ├→ PROPOSER sub-agent    → returns proposal
      ├→ SPEC WRITER sub-agent → returns specs
      ├→ DESIGNER sub-agent    → returns design
      ├→ TASK PLANNER          → returns tasks
      ├→ IMPLEMENTER           → returns code
      ├→ VERIFIER              → returns report
      └→ ARCHIVER              → returns archive

Persistence Modes

All commands respect the configured artifact store mode:

• engram (recommended) - Artifacts stored in Engram. Zero project files.
• openspec - Artifacts in openspec/ directory. Full file-based workflow.
• none - Ephemeral artifacts. Nothing persisted between sessions.

See Artifact Storage for details.

Getting Started

Start with the initialization command:

Initialize SDD

Run /sdd-init to bootstrap SDD in your project