Antigravity Kit

AI Agent Capability Expansion Toolkit - Bộ công cụ mở rộng khả năng AI coding assistant với 60+ skills, rules, và workflows chuyên biệt.

🌟 Introduction

Antigravity Kit là bộ công cụ toàn diện bao gồm:

Skills - Chuyên môn domain-specific (React, Node.js, Database, Testing, UI/UX...)
Rules - Guidelines và constraints cho agent behavior
Modes - Task-specific workflows (Build, Debug, Optimize, Review...)
Commands - Slash commands để thực hiện tác vụ nhanh chóng
Agents - Specialized agents cho development tasks

Key Features

✅ 68 Skills - Lazy-loaded để tiết kiệm context
✅ 14 Commands - Nhanh chóng thực hiện tác vụ phức tạp
✅ 13 Rules - Hướng dẫn agent behavior
✅ Dual IDE Support - Hoạt động trên cả Antigravity và OpenCode
✅ Enhanced Workflow - 3-level debugging, TDD workflow, security checklist
✅ Vibecode Integration - Partnership model cho project development

Powered By

UI UX Pro Max - Design Intelligence với 50 styles, 21 palettes, 50 font pairings
ClaudeKit - Production-ready AI subagents và workflows

🚀 What's New in v4.1.1

Codex Skill Naming Update

Renamed custom Codex skills to shorter IDs (removed vibecode- prefix).
Examples: mode-vision, mode-blueprint, template-saas-app, quality-gates.

New Codex Usage Documentation

Added detailed per-skill guide:
- templates/.codex/SKILLS_USAGE_GUIDE.md
Includes: when-to-use matrix, per-skill prompts, lifecycle combos, troubleshooting.

Template and Config Alignment

Updated template skill indexes under:
- templates/.codex/skills/index.md
- templates/.agents/skills/index.md
Updated config path examples in:
- templates/.codex/config.toml

Previous Major Additions (v4.0)

Vibecode lifecycle commands: /vision, /blueprint, /investigate, /qa, /xray, /handover.
Partnership workflow for planning-first project execution.

🎯 Supported IDEs

IDE	Folder	Features	Context Usage
OpenCode	`.opencode/` + `AGENTS.md`	Skills loaded on-demand, minimal context	~2500 tokens
Antigravity	`.agent/`	Full rules với activation types, workflows	Standard

Both IDEs share cùng skills và capabilities, optimized cho mỗi platform's architecture.

📦 Installation

Using CLI (Recommended)

# Navigate to your project
cd your-project

# Install cho both IDEs (recommended)
npx @duck4nh/antigravity-kit init

# Or specify target IDE
npx @duck4nh/antigravity-kit init --ide opencode
npx @duck4nh/antigravity-kit init --ide antigravity
npx @duck4nh/antigravity-kit init --ide both

CLI Commands

Command	Description
`npx @duck4nh/antigravity-kit init`	Install vào current directory
`npx @duck4nh/antigravity-kit update`	Update đến latest version
`npx @duck4nh/antigravity-kit status`	Check installation status

CLI Options

Option	Description
`-i, --ide <target>`	Target IDE: `antigravity`, `opencode`, hoặc `both`
`-f, --force`	Overwrite existing installation
`-p, --path <dir>`	Custom project path
`-b, --branch <name>`	Use specific branch

🎮 Quick Start

For OpenCode Users

# 1. Install
npx @duck4nh/antigravity-kit init --ide opencode

# 2. Skills tự động load khi needed
User: "Fix bug in this React component"
Agent: (automatically loads react-expert skill)

# 3. Use slash commands
User: /fix "Login button not working"
User: /review src/
User: /test

# 4. Press Tab to switch giữa Build và Plan modes

For Antigravity Users

# 1. Install
npx @duck4nh/antigravity-kit init --ide antigravity

# 2. Rules tự động apply dựa trên activation type
# - always_on: Luôn active
# - model_decision: Agent quyết định khi apply

# 3. Skills có sẵn trong .agent/skills/
# 4. Workflows có sẵn trong .agent/workflows/

💻 Commands

Core Commands

Command	Purpose	Usage	Time
`/fix`	3-Level Tiered Debugging	`/fix "Bug description" [--level 1\|2\|3]`	5-60 min
`/review`	Structured Code Review	`/review <path>`	10-15 min
`/test`	Test Execution	`/test [test-pattern]`	2-5 min
`/commit`	Smart Git Commit	`/commit "message"`	1 min
`/clean`	Code Cleanup	`/clean <path>`	5-10 min
`/enhance`	Code Enhancement	`/enhance <path>`	10-15 min
`/plan`	Planning Mode	`/plan "task"`	5-10 min
`/think`	Deep Analysis	`/think "question"`	5-10 min

Vibecode Commands (Partnership Model)

Command	Purpose	When to Use
`/vision`	AI proposes vision first	Starting new project from scratch
`/blueprint`	Create detailed blueprint	After vision is approved
`/investigate`	Deep debugging protocol	Complex bugs cần thorough investigation
`/qa`	Quality assurance planning	Creating comprehensive test plans
`/xray`	Project documentation	Handover, onboarding, codebase analysis
`/handover`	Alias cho /xray	Same as /xray

Command Comparison

Task	Old Way	New Way (v4.1.1)	Benefit
Simple bug	`/fix` + manual steps	`/fix` (Level 1 auto)	⚡ 5-10 min
Complex bug	Manual investigation	`/investigate` hoặc `/fix --level 3`	🔬 Thorough analysis
New project	Start coding immediately	`/vision` → `/blueprint` → build	🎯 Better architecture
Code review	Manual checklist	`/review` (structured output)	📊 Actionable feedback

🎓 Skills

Skills are domain-specific expertise modules loaded on-demand bởi agent.

Frontend

Skill	Description	Triggers
`react-expert`	React 18/19, hooks, patterns, performance	react, component, hooks
`nextjs-expert`	App Router, Server Components, SSR	next, nextjs, app router
`css-expert`	CSS architecture, responsive, design systems	css, styling, responsive
`state-management-expert`	Redux, Zustand, React Query	state, redux, store
`ui-ux-pro-max`	50 styles, 21 palettes, 50 font pairings	design, ui, ux, styling
`accessibility-expert`	WCAG compliance, a11y testing	accessibility, a11y, wcag

Backend

Skill	Description	Triggers
`nodejs-expert`	Async patterns, modules, performance	node, nodejs, backend
`nestjs-expert`	Module architecture, DI, testing	nest, nestjs
`rest-api-expert`	RESTful design, HTTP semantics	api, rest, endpoint
`auth-expert`	JWT, OAuth 2.0, RBAC, security	auth, authentication, jwt

Database

Skill	Description	Triggers
`prisma-expert`	Schema design, migrations, queries	prisma, orm, schema
`database-expert`	General database optimization	database, db, sql
`postgres-expert`	PostgreSQL-specific patterns	postgres, postgresql
`mongodb-expert`	MongoDB document modeling	mongodb, mongo
`redis-expert`	Caching, sessions, rate limiting	redis, cache

Testing

Skill	Description	Triggers
`testing-expert`	General testing strategies	test, testing, tdd
`jest-expert`	Jest framework, mocking	jest, unit test
`vitest-expert`	Vitest, Vite integration	vitest
`playwright-expert`	E2E testing, browser automation	playwright, e2e

DevOps & Tools

Skill	Description	Triggers
`devops-expert`	CI/CD, infrastructure	devops, cicd, deployment
`docker-expert`	Containerization, Compose	docker, container
`github-actions-expert`	GitHub Actions workflows	github actions, workflow
`git-expert`	Git workflows, conflicts	git, merge, conflict
`typescript-expert`	TypeScript patterns, type system	typescript, ts, types

Workflow Modes

Skill	Description	Triggers
`mode-consulting`	Compare options, make recommendations	should, recommend, compare
`mode-build`	Create features, components, modules	create, build, implement
`mode-debug`	Fix bugs, analyze root cause	fix, debug, error
`mode-optimize`	Refactor, improve performance	optimize, refactor, improve
`mode-learn`	Explain concepts, code, architecture	explain, how does, what is
`mode-review`	Code review, security audit	review, check, audit
`mode-migrate`	Version upgrades, stack migrations	upgrade, migrate, update
`mode-investigate`	Deep debugging protocol	investigate, deep debug
`mode-qa`	Quality assurance planning	qa, test plan, quality
`mode-xray`	Project documentation	xray, handover, onboarding

Security & Operations

Skill	Description	Triggers
`pentest-expert`	Penetration testing methodology	pentest, security test
`web-security-expert`	OWASP, XSS, SQLi, CSRF	security, vulnerability
`exploit-dev-expert`	Exploit development, PoC	exploit, poc, payload
`mode-tool-dev`	Security tool development	script, automate, scanner
`cloud-deployment-expert`	VPS, AWS, Docker deployment	deploy, production, vps
`linux-server-expert`	Ubuntu/Debian, Nginx, SSL	server, nginx, ssl
`system-design-expert`	Architecture, scalability	architecture, design, scale

📋 Rules

Rules là guidelines điều hướng agent behavior (Antigravity only).

Activation Types

Type	Description
`always_on`	Luôn được áp dụng
`model_decision`	Agent tự quyết định khi nào áp dụng

Rules List

File	Activation	Description
`01-identity.md`	Always On	Role & working principles
`02-task-classification.md`	Always On	Task types classification
`03-mode-consulting.md`	Model Decision	Consulting process
`04-mode-build.md`	Model Decision	Build process
`05-mode-debug.md`	Model Decision	Debug process
`06-mode-optimize.md`	Model Decision	Optimization process
`07-mode-learn.md`	Model Decision	Learning/explaining process
`08-mode-review.md`	Model Decision	Code review process
`09-mode-migrate.md`	Model Decision	Migration process
`10-technical-standards.md`	Always On	Coding standards
`11-communication.md`	Always On	Communication style
`12-checklist.md`	Always On	Pre-delivery checklist
`13-special-situations.md`	Always On	Edge case handling

🤖 Agents

Core OpenCode Agents

Agent	Triggers	Expertise
`fullstack-dev`	frontend, backend, react, api	Next.js, React, Node.js, APIs
`infra-expert`	database, devops, docker, deploy	PostgreSQL, MongoDB, Docker, CI/CD
`code-reviewer`	review, audit, check	Code quality, architecture, best practices
`security-expert`	security, vulnerability	OWASP, secure coding, vulnerability assessment

Vibecode Agents (Partnership Model)

Agent	Triggers	Role	Description
Architect	vision, propose, design	Leads with patterns	Proposes complete vision first
Builder	build, implement	Executes vision	Implements approved blueprint
Investigator	investigate, debug, analyze	Deep debugging	9-step investigation protocol
QA Specialist	qa, test planning, quality	Quality assurance	Tiered testing approach
X-Ray	xray, handover, document	Documentation	Project analysis & handover

💡 Partnership Model: AI proposes vision first (80% patterns), you provide context (20% unique) = Perfect product

📚 Usage Examples

Example 1: Quick Bug Fix (Level 1)

# Agent auto-classifies as Level 1 (simple bug)
User: /fix "Email validation not working"

Agent Output:
🔍 Analyzing bug...
📊 Bug Classification: Level 1 (Simple)
⚡ Quick Fix Workflow initiated (5-10 min)

1. Reproduce → Enter invalid email, nothing happens
2. Diagnose → Missing regex validation
3. Fix → Add email validation regex
4. Verify → Test with valid/invalid emails

✅ Fixed: Email validation now works
✅ Tested: Valid emails pass, invalid emails show error
⏱️ Time: 8 minutes

Example 2: Partnership Model - New Project

# Step 1: AI proposes vision
User: /vision "I need a landing page for my SaaS product"

Agent: Proposes complete architecture:
- Hero section with CTA
- Social proof (logos, testimonials)
- Problem/Solution framework
- Pricing table
- FAQ section
- Recommended tech stack: Next.js + Tailwind + shadcn/ui

# Step 2: Provide context
User: "Our audience is developers. Focus on API features and code examples."

Agent: Adjusts vision to include:
- Interactive API demo
- Code snippets in multiple languages
- Developer-focused copy

# Step 3: Create blueprint
User: /blueprint

Agent: Creates detailed blueprint with all sections, components, and implementation plan

Example 3: Deep Investigation (Level 3)

User: /fix "Production API timeout at 2-3 PM daily" --level 3

Agent Process:
1. Pre-Debug: Gather info (logs, recent changes, metrics)
2. Reproduce: Load test at peak traffic
3. Diagnose: API times, DB queries, memory profiling
4. Hypothesize: 5 potential causes
5. Confirm: API timeout + no retry + slow DB query
6. Fix: Timeout + Retry + Circuit breaker
7. Verify: Load test, performance comparison
8. Impact: p95 -90%, errors -93%
9. Rollback: Document rollback triggers & steps

✅ Fixed: API handles peak traffic
📊 Performance: p95 800ms (was 8000ms), error rate 0.1% (was 15%)
⏱️ Time: 45 minutes

Example 4: Code Review

User: /review src/services/auth.ts

Agent Output:
📊 Code Review Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📈 Metrics
  Files: 1
  Lines: 245
  Functions: 12
  Complexity: Medium

🚨 Critical Issues (2)
  [!] Hardcoded JWT secret in line 45
  [!] No rate limiting on login endpoint

⚠️ Major Issues (3)
  [*] Missing error handling in refreshToken()
  [*] No input validation on email field
  [*] Token expiration not configurable

💪 Strengths
  ✓ Clean separation of concerns
  ✓ Good use of TypeScript types
  ✓ Comprehensive JSDoc comments

📋 Action Items
  1. Move JWT secret to environment variables
  2. Add rate limiting middleware
  3. Implement error boundaries
  4. Add input validation with Zod

📁 Project Structure

your-project/
├── .opencode/                  # OpenCode configuration
│   ├── opencode.json           # Config file
│   ├── skill/                  # 60+ skills (lazy-loaded)
│   │   ├── react-expert/
│   │   ├── typescript-expert/
│   │   └── ...
│   ├── command/                # 14+ slash commands
│   │   ├── fix.md
│   │   ├── review.md
│   │   ├── vision.md
│   │   └── ...
│   ├── agent/                  # Specialized agents
│   │   ├── fullstack-dev.md
│   │   ├── security-expert.md
│   │   └── ...
│   └── instructions/           # Workflow instructions
│       ├── pre-delivery-checklist.md
│       ├── security-checklist.md
│       ├── tdd-workflow.md
│       └── error-handling-advanced.md
├── .agent/                     # Antigravity configuration
│   ├── rules/                  # 13 rule files
│   │   ├── 01-identity.md
│   │   ├── 02-task-classification.md
│   │   └── ...
│   ├── skills/                 # 35+ skills
│   │   ├── react-expert/
│   │   ├── nestjs-expert/
│   │   └── ...
│   └── workflows/              # Workflow definitions
│       ├── request.md
│       └── ui-ux-pro-max.md
├── shared/                     # Shared resources
│   └── ui-ux-pro-max/          # UI/UX design data
│       ├── data/               # CSV databases
│       │   ├── colors.csv
│       │   ├── styles.csv
│       │   └── ...
│       └── scripts/            # Search scripts
├── .vibecode/                  # Vibecode integration
│   ├── README.md               # Vibecode philosophy
│   ├── PHILOSOPHY.md           # Partnership model
│   └── workflows/              # MASTER workflows
│       ├── VIBECODE-MASTER.md
│       ├── DEBUG-MASTER.md
│       └── ...
└── AGENTS.md                   # Core rules cho OpenCode

🎓 Credits

Project này được xây dựng dựa trên và inspired bởi:

Project	Description	Link
UI UX Pro Max	Design Intelligence - 50 styles, 21 palettes, 50 font pairings	ui-ux-pro-max-skill.nextlevelbuilder.io
ClaudeKit	Production-ready AI subagents và workflows	claudekit.cc
Vibecode Kit	Partnership model workflows & templates	vibecode.cc

🤝 Contributing

Adding a New Skill

Create folder in both locations:
- .opencode/skill/your-skill/SKILL.md
- .agent/skills/your-skill/SKILL.md
Use correct frontmatter:

OpenCode format:

---
name: your-skill
description: >-
  Skill description (max 1024 chars). Explain when to use this skill.
license: MIT
compatibility: opencode
metadata:
  category: frontend
---

# Your Skill

Instructions for the agent...

Antigravity format:

---
name: your-skill
description: Skill description. Use when X or Y.
---

# Your Skill

Instructions for the agent...

Adding a New Command

Create file: .opencode/command/your-command.md
Add frontmatter:

---
description: Command description
argument-hint: "[optional-arg]"
---

# /your-command

Command purpose and usage...

Adding a New Rule (Antigravity only)

Create file: .agent/rules/XX-your-rule.md
Add frontmatter:

---
activation: always_on | model_decision
description: When to apply (if model_decision)
---

# Your Rule

Content...

📄 License

MIT License - See LICENSE for details.

Made with ❤️ by duck4nh

⭐ Star us on GitHub!

Name		Name	Last commit message	Last commit date
Latest commit History 22 Commits
.github/workflows		.github/workflows
bin		bin
templates		templates
.gitignore		.gitignore
.npmignore		.npmignore
CHANGELOG.md		CHANGELOG.md
README.md		README.md
VIBECODE_INTEGRATION_TASKS.md		VIBECODE_INTEGRATION_TASKS.md
package-lock.json		package-lock.json
package.json		package.json

duck4nh/antigravity-kit

Folders and files

Latest commit

History

Repository files navigation