Skip to main content

Quick Start Guide

Get up and running with Antigravity Kit in just a few minutes. This guide walks you through installation, configuration, and your first agent interaction.

Getting Started

1

Install Antigravity Kit

Install the .agent folder into your project using npx:
npx @vudovn/ag-kit init
This creates a .agent folder containing:
  • 20 specialist agents
  • 37 domain-specific skills
  • 11 slash command workflows
  • Validation scripts and documentation
2

Configure Git (Optional)

If you want to keep .agent local without tracking it in Git:
echo ".agent/" >> .git/info/exclude
Do NOT add .agent/ to .gitignore if you’re using Cursor or Windsurf. This prevents the IDE from indexing workflows and showing slash command suggestions.
3

Open Your AI Editor

Launch Cursor or Windsurf and open your project. The editor will automatically index the .agent folder and make agents and workflows available.
4

Start Using Agents

Simply describe what you need - agents are automatically selected:
You: "Add JWT authentication to the login endpoint"
AI: 🤖 Applying @security-auditor + @backend-specialist...
The AI will:
  • Detect relevant domains (security + backend)
  • Load appropriate skills (api-patterns, nodejs-best-practices)
  • Provide expert-level implementation

Using Agents

Automatic Selection

Agents are automatically selected based on your request. No need to explicitly mention them! Examples:
You: "Fix the dark mode button"
AI: 🤖 Using @frontend-specialist...
     Skills: frontend-design, react-best-practices

You: "Optimize database queries"
AI: 🤖 Using @database-architect...
     Skills: database-design, prisma-expert

You: "Login returns 500 error"
AI: 🤖 Using @debugger for systematic analysis...
     Skills: systematic-debugging

Available Agents

AgentWhen It’s UsedSkills Loaded
frontend-specialistUI/UX work, React componentsreact-best-practices, tailwind-patterns
backend-specialistAPI development, server logicapi-patterns, nodejs-best-practices
database-architectSchema design, SQL queriesdatabase-design, prisma-expert
security-auditorSecurity compliance, auditsvulnerability-scanner
debuggerBug investigation, errorssystematic-debugging
test-engineerWriting tests, test strategiestesting-patterns, webapp-testing
devops-engineerDeployment, CI/CDdocker-expert, deployment-procedures
performance-optimizerSpeed optimizationperformance-profiling
View all 20 agents in .agent/agents/ or check the ARCHITECTURE.md file.

Using Slash Commands

Workflows are invoked using slash commands in your AI editor. Type / to see available commands.

Common Workflows

Explore options before implementation:
/brainstorm authentication system for multi-tenant app
The AI will:
  • Ask strategic questions
  • Explore different approaches
  • Provide recommendations
  • Create a structured discovery document

All Available Workflows

CommandDescriptionUse Case
/brainstormSocratic discoveryUnclear requirements, need options
/createCreate features/appsNew functionality
/debugSystematic debuggingBug reports, errors
/deployDeploy applicationReady to ship
/enhanceImprove existing codeRefactoring, optimization
/orchestrateMulti-agent coordinationComplex full-stack tasks
/planTask breakdownLarge projects, estimates
/previewPreview changes locallyBefore deploying
/statusCheck project statusHealth checks
/testGenerate and run testsBefore deployment
/ui-ux-pro-maxDesign with 50 stylesProfessional UI design

Real-World Examples

Example 1: Building a Feature

You: /create user profile page with avatar upload

AI: 🤖 Using @frontend-specialist...
    Creating:
    - app/profile/page.tsx (Next.js server component)
    - components/AvatarUpload.tsx (React component)
    - lib/upload-utils.ts (utility functions)
    
    Applied skills:
    ✓ react-best-practices (server components, hooks)
    ✓ tailwind-patterns (responsive design)
    ✓ frontend-design (UX patterns)

Example 2: Debugging an Issue

You: /debug API returns 500 when creating a post

AI: 🤖 Using @debugger...
    Analyzing:
    1. Checking error logs... Found: "Cannot read property 'id' of undefined"
    2. Root cause: Missing user authentication check
    3. Location: app/api/posts/route.ts:15
    
    Fix:
    [Code example provided]
    
    Verification:
    ✓ Added auth middleware
    ✓ Added error handling
    ✓ Suggested test cases

Example 3: Multi-Agent Task

You: /orchestrate build e-commerce product page with cart

AI: 🤖 Orchestrating multiple specialists...
    
    @frontend-specialist:
    - Building ProductPage component
    - Creating Cart UI with animations
    
    @backend-specialist:
    - API routes for cart operations
    - Session management
    
    @database-architect:
    - Cart schema design
    - Product inventory tracking
    
    @test-engineer:
    - E2E cart flow tests
    - Unit tests for calculations

Validation Scripts

Antigravity Kit includes validation scripts to ensure code quality:
python .agent/scripts/checklist.py .
checklist.py (Development - ~30 seconds):
  • Security scan
  • Code quality (lint, types)
  • Schema validation
  • Unit tests
  • UX audit
  • SEO check
verify_all.py (Pre-deployment - ~3-5 minutes):
  • All quick checks
  • Lighthouse (Core Web Vitals)
  • Playwright E2E tests
  • Bundle analysis
  • Mobile audit
  • i18n check

Tips for Best Results

Be Specific

Provide clear requirements and context for better results

Use Workflows

Use / commands for structured, multi-step processes

Trust Auto-Selection

Let the system choose agents - it’s trained to detect domains

Run Validations

Use validation scripts before committing or deploying

Next Steps

Now that you’re familiar with the basics:
  • Explore the .agent/agents/ folder to see all 20 specialist agents
  • Check out .agent/skills/ to discover 37 domain-specific skills
  • Read .agent/ARCHITECTURE.md for detailed system documentation
  • Try different workflows with / commands in your AI editor

View Full Documentation

Browse comprehensive guides and examples online

Build docs developers (and LLMs) love