/agentful-init

Interactive onboarding wizard that guides you through agentful setup with 7 questions.

Overview

The /agentful-init command provides a guided, interactive setup experience for first-time agentful users. It asks 7 focused questions to understand your project and automatically generates your product specification.

When to use:

First time using agentful
New project with no product spec
Want guided setup instead of manual editing

What it creates:

.claude/product/index.md - Your product specification
Auto-triggers /agentful-generate to create specialized agents
.agentful/setup-progress.json - Tracks setup state

Usage

/agentful-init

The command will:

Check if already initialized (offers to continue or restart)
Auto-detect your tech stack
Ask 7 guided questions
Generate product specification
Automatically generate specialized agents
Prepare for /agentful-start

The 7 Questions

Question 1: Product Description

What are you building?
 
Describe your product in 1-3 sentences:
- What problem does it solve?
- Who is it for?
- What makes it unique?

Example: "A task management app for remote teams that integrates with Slack and automatically prioritizes work based on deadlines and dependencies."

Question 2: Tech Stack Confirmation

I detected this tech stack:
- Language: TypeScript
- Framework: Next.js 15
- Database: PostgreSQL with Prisma
- Testing: Vitest
 
Is this correct?

Options:

Yes, correct
No, let me specify manually

Question 3: Key Features

What are the main features you want to build?
 
List 3-10 key features, one per line.

Example:

User authentication with email/password
Task creation and assignment
Real-time notifications
Calendar integration
Team collaboration spaces

Question 4: Feature Priorities

Should I:
- Auto-assign priorities based on typical MVP (recommended)
- Let me set priorities for each feature manually

Auto-assignment logic:

First 30% of features → CRITICAL
Next 40% → HIGH
Remaining → MEDIUM

Question 5: Testing Strategy

What level of testing do you want?
- Comprehensive (unit + integration + E2E)
- Standard (unit + integration)
- Minimal (unit tests only)

Sets coverage target:

Comprehensive: 85%
Standard/Minimal: 80%

Question 6: Deployment Target

Where will this be deployed?
- Vercel
- AWS (EC2, Lambda, etc.)
- Google Cloud
- Docker/Self-hosted
- Not sure yet / Multiple targets

Helps optimize build configs and suggest best practices.

Question 7: Experience Level

What's your experience level with [detected framework]?
- Expert - I know the internals
- Proficient - I build with it regularly
- Intermediate - I know the basics
- Beginner - First time using this stack

Adjusts verbosity of agent explanations:

Expert → Concise
Proficient/Intermediate → Moderate
Beginner → Detailed

Generated Product Spec

The command creates .claude/product/index.md with this structure:

# [Your Product Name]
 
[Product description from Q1]
 
## Tech Stack
 
- **Language**: TypeScript
- **Framework**: Next.js 15
- **Database**: PostgreSQL with Prisma
- **Testing**: Vitest
- **Deployment**: Vercel
 
## Features
 
### 1. User Authentication - CRITICAL
[Auto-generated description based on feature name]
 
**Acceptance Criteria**:
- [ ] Email/password registration
- [ ] Login with session management
- [ ] Password reset flow
- [ ] Email verification
 
**Priority**: CRITICAL
**Status**: Not started
 
---
 
### 2. Task Creation - CRITICAL
...
 
[Continues for all features]

Resume Support

If interrupted, /agentful-init can resume where you left off:

Found incomplete setup from 2026-01-25T10:30:00Z.
 
Completed so far:
✓ Product description
✓ Tech stack
✓ Feature list
✗ Priority assignment
 
Would you like to:
- Resume where I left off
- Start over from beginning

Progress is saved to .agentful/setup-progress.json.

After Completion

Once setup completes, you'll see:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
   ✅ Setup Complete!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
Your agentful project is ready for autonomous development.
 
What was created:
  📄 .claude/product/index.md (8 features)
  🤖 Specialized agents (generated for your domains)
  📚 Tech skills (Next.js, Prisma, etc.)
  ⚙️  Quality gates (tests, linting, security)
 
Next steps:
 
1. Review your product spec:
   cat .claude/product/index.md
 
2. Start development:
   /agentful-start
 
3. Monitor progress:
   /agentful-status
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

vs /agentful-product

Feature	/agentful-init	/agentful-product
Best for	First-time users	Existing projects
Interaction	7 guided questions	Auto-detects workflow
Creates spec	Yes	Yes (if missing)
Analyzes spec	No	Yes
Refinement mode	No	Yes
Q&A mode	No	Yes
Tech detection	Yes	Yes

Recommendation: Use /agentful-init for your first project, then use /agentful-product for analysis and refinement.

Comparison with Other Commands

# For brand new users
/agentful-init         # Start here - guided setup
 
# After setup is complete
/agentful-start        # Begin development (auto-generates agents on first run)
 
# To analyze existing spec
/agentful-product      # Check readiness, find gaps
 
# To check if setup is healthy
/agentful-analyze      # Validate configuration

Troubleshooting

Setup fails mid-way → Run /agentful-init again - it will offer to resume

Want to start over → Delete .agentful/setup-progress.json, then run /agentful-init

Already initialized, ran by mistake → Choose "Continue with existing setup" when prompted

Questions too basic/advanced → Your answer to Q7 (experience level) adjusts future agent verbosity