Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

agentful Commands Reference

Complete reference guide for all agentful slash commands used in autonomous product development.

Overview

agentful provides a set of slash commands that automate the product development lifecycle. These commands work together to create an autonomous development loop that can build features, validate quality, and resolve decisions with minimal human intervention.

Command Categories

Core Development Commands

Quick Start

Basic Workflow

The typical development workflow follows this sequence:

1. /agentful-start     → Begin autonomous development
2. /agentful-status    → Check progress
3. /agentful-decide    → Answer blocking decisions (if needed)
4. /agentful-validate  → Verify quality (optional)
5. /agentful-start     → Continue development

First Time Setup

Before using agentful commands, ensure you have:

  1. PRODUCT.md - Product requirements document
  2. .agentful/ directory - State management
  3. Git repository - Version control
# Initialize agentful
mkdir -p .agentful
echo '{}' > .agentful/state.json
echo '{}' > .agentful/completion.json
echo '{"pending":[],"resolved":[]}' > .agentful/decisions.json

Command Reference

/agentful-start

Purpose: Initiates the autonomous development loop

When to use:
  • Starting a new product development session
  • Resuming development after a break
  • Continuing after resolving decisions
  • Running autonomous development cycles
Key features:
  • Reads product state and picks next task
  • Delegates work to specialist agents
  • Validates completion before moving on
  • Updates completion tracking

See: Complete /agentful-start reference

/agentful-status

Purpose: Shows current progress and development state

When to use:
  • Checking what's been completed
  • Seeing what's currently in progress
  • Identifying blocking issues
  • Getting an overview before making decisions
Key features:
  • Visual progress display with percentages
  • Feature status table (Done/Active/Pending)
  • Quality gate status
  • Pending decisions overview
  • Current work context

See: Complete /agentful-status reference

/agentful-decide

Purpose: Resolves pending decisions blocking development

When to use:
  • /agentful-start reports pending decisions
  • You need to make architectural choices
  • Blocking issues need resolution
  • Setting product direction
Key features:
  • Interactive decision interface
  • Presents options with context
  • Records decisions for traceability
  • Unblocks features automatically

See: Complete /agentful-decide reference

/agentful-validate

Purpose: Runs quality checks and validation gates

When to use:
  • Before committing code
  • After completing features
  • Verifying quality standards
  • CI/CD integration
Key features:
  • TypeScript type checking
  • Lint validation
  • Dead code detection
  • Test execution
  • Coverage reporting
  • Security scanning

See: Complete /agentful-validate reference

Common Workflows

Workflow 1: Initial Development Session

# Start building your product
/agentful-start
 
# Check progress after some time
/agentful-status
 
# If blocked by decisions
/agentful-decide
 
# Continue development
/agentful-start

Workflow 2: Quality-Focused Development

# Run development
/agentful-start
 
# Validate quality before proceeding
/agentful-validate
 
# If validation fails, run again to auto-fix
/agentful-start
 
# Verify fixes
/agentful-validate

Workflow 3: Autonomous Loop (Advanced)

For continuous autonomous development:

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

This runs the development loop continuously until:

  • All features complete (100%)
  • All quality gates passing
  • Maximum iterations reached

Workflow 4: Decision-Heavy Session

# Start development
/agentful-start
 
# If blocked, review all pending decisions
/agentful-decide
 
# Resolve multiple decisions in sequence
# (Interactive mode walks through each)
 
# Resume with clear path
/agentful-start
 
# Verify unblocked progress
/agentful-status

Command Output Reference

Status Indicators

SymbolMeaning
✅Completed/Passing
🔄In Progress
⏸Pending/Blocked
⚠️Warning
❌Failed/Error

Progress Bars

Overall Progress: ████░░░░░░░░░░ 48%
                  10%     50%     100%
  • Solid blocks (█) = completed
  • Empty blocks (░) = remaining

Quality Gates

┌─────────────────────┬────────┐
│ Quality Gate        │ Status │
├─────────────────────┼────────┤
│ Tests Passing       │ ✅     │
│ No Type Errors      │ ✅     │
│ No Dead Code        │ ❌     │
│ Coverage ≥ 80%      │ ⚠️ 72% │
└─────────────────────┴────────┘

Tips and Best Practices

Development Tips

  1. Start with /agentful-status - Always check current state before starting work
  2. Resolve decisions promptly - Pending decisions block progress immediately
  3. Run validation regularly - Catch quality issues early
  4. Monitor iterations - High iteration counts may indicate problems

Productivity Tips

  1. Use autonomous mode for long sessions - Let the loop run while you focus on other tasks
  2. Check status between iterations - Stay informed without interrupting flow
  3. Keep PRODUCT.md updated - Better requirements = better autonomous development
  4. Review decisions log - Learn from previous architectural choices

Troubleshooting

Development Stuck?

# Check what's happening
/agentful-status
 
# Look for pending decisions
# If present, resolve them
/agentful-decide

Quality Gates Failing?

# Run detailed validation
/agentful-validate
 
# Review specific failures
# Run /agentful-start to auto-fix

Can't Start Development?

# Verify required files exist
ls -la PRODUCT.md .agentful/
 
# Initialize if missing
mkdir -p .agentful
echo '{"version":"1.0","current_task":null}' > .agentful/state.json

Advanced Usage

Integration with CI/CD

# .github/workflows/agentful.yml
name: agentful Validation
on: [push, pull_request]
jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v2
      - name: Run agentful validation
        run: |
          npm install
          /agentful-validate

Custom Decision Templates

Create reusable decision patterns:

// .agentful/decision-templates.json
{
  "templates": [
    {
      "name": "auth-strategy",
      "question": "Authentication approach?",
      "options": ["JWT", "Sessions", "OAuth", "Custom"]
    }
  ]
}

Progress Tracking Integration

# Sync with external project management
/agentful-status | jq '.features' > project-status.json

See Also