Commands Reference

Complete reference for agentful slash commands.

Overview

Agentful provides slash commands for structured product development. These commands automate the development lifecycle with human checkpoints for key decisions.

Command Categories

Planning Commands

/agentful-product - Create and refine product requirements
/agentful-init - Interactive onboarding with 7 guided questions

Generation Commands

/agentful-generate - Analyze codebase, generate domain agents and tech skills
/agentful-update - Smart update mechanism - updates configuration while preserving customizations

Core Development Commands

/agentful-start - Start or resume development
/agentful-status - View current progress
/agentful-decide - Resolve pending decisions
/agentful-validate - Run quality checks

Analysis Commands

/agentful-analyze - Detect gaps and validate configuration

Conversational Interface

/agentful - Natural language interface for all commands

Quick Start

Basic workflow:

# 1. Define requirements (choose one):
/agentful-init      # Interactive 7-question wizard (recommended for new users)
# OR
/agentful-product   # Manual spec creation/analysis
 
# 2. Start development (auto-generates agents on first run)
/agentful-start
 
# 3. Check progress
/agentful-status
 
# 4. Resolve decisions if needed
/agentful-decide
 
# 5. Run quality checks
/agentful-validate

Note: You don't need to manually run /agentful-generate - it runs automatically on first /agentful-start. Only re-run /agentful-generate manually if you significantly change your tech stack.

Setup

Before using commands, ensure you have:

Product specification - .claude/product/index.md
.agentful/ directory - State management
Git repository - Version control

Initialize:

mkdir -p .agentful
echo '{}' > .agentful/state.json
echo '{}' > .agentful/completion.json
echo '{"pending":[],"resolved":[]}' > .agentful/decisions.json

Common Workflows

Workflow 1: Initial Development

/agentful-init        # Interactive setup (or /agentful-product for manual)
/agentful-start       # Auto-generates agents, then begins development
/agentful-status      # Check progress

Workflow 2: Ongoing Development

/agentful-start       # Continue work
/agentful-status      # Check progress
/agentful-decide      # Resolve decisions if blocked
/agentful-validate    # Run quality checks

Workflow 3: Extended Sessions

/ralph-loop "/agentful-start" --max-iterations 50 --completion-promise "AGENTFUL_COMPLETE"

Runs development loop with pauses for:

Human decisions
All features complete (100%)
Quality gates passing
Maximum iterations reached

Tips

Start with /agentful-status - Check state before starting work
Resolve decisions promptly - Pending decisions block progress
Run validation regularly - Catch quality issues early
Keep product spec updated - Better requirements = better development
Monitor iterations - High count may indicate issues