Agent Skills CLI (@uocnv1998/agent-kit)

A powerful command-line tool designed to scaffold and manage AI Agent environments. This tool specifically builds the foundation for AI agents (like Antigravity) by setting up structured Skills, Rules, and Workflows within your project.

🚀 Overview

The Agent Skills CLI streamlines the process of initializing a project for specific AI roles. It ensures that your AI collaborator has the right context, tools, and operational procedures to be effective immediately.

🛠 Installation & Usage

You can run this tool directly using npx without installation:

npx @uocnv1998/agent-kit --role=<role_name>

Replace <role_name> with one of the supported roles.

Examples:

# Standard role installation
npx @uocnv1998/agent-kit --role=dev

# DA role with interactive AI framework selection
npx @uocnv1998/agent-kit --role=da
# → Choose between: Antigravity (default) or Claude

🎭 Supported Roles

| Role | Description | | :--- | :--- | | po_ba | Product Owner / Business Analyst context and requirements gathering skills. | | designer | UI/UX design patterns and aesthetic guidelines. | | dev | Development standards, coding patterns, and technical implementation skills. | | tester | Quality assurance, testing strategies, and bug reporting workflows. | | da | NEW Data Analytics workflow with multi-agent coordination (choose Antigravity or Claude). | | all | Installs the full suite of roles and global configurations. |

📁 Project Structure

When you run the command, it creates or updates a .agent/ directory in your project root:

your-project/
├── .agent/
│   ├── rules/       # Global and role-specific operational rules
│   ├── skills/      # Specialized instructions for different agent roles
│   └── workflows/   # Step-by-step guides for common tasks
└── ...

⚙️ How it Works

The CLI copies pre-defined templates from the @uocnv1998/agent-kit package into your local .agent folder:

1. Selective Skills: Only the skills relevant to the chosen role are copied.
2. Workflows: Role-specific *-workflow.md files are placed in the workflows directory.
3. Rules: Role-specific rules are established to guide the agent's behavior.

🤖 AI Framework Selection (DA Role Only)

For the Data Analytics (da) role, you'll be prompted to choose your AI framework:

? Which AI framework would you like to use?
  ◉ Antigravity - Default Antigravity framework
  ○ Claude - Claude AI with advanced features (Opus 4.6, Sonnet 4.6)

Use arrow keys ↑↓ to navigate, SPACE/ENTER to select.

Antigravity Setup

• Agents use model: inherit (flexible)
• Rules: .agent/rules/GEMINI.md
• Minimal overhead, maximum flexibility

Claude Setup

• Agents pre-optimized with Opus 4.6 and Sonnet 4.6
• Rules: .agent/rules/CLAUDE.md (advanced optimization)
• Features: Prompt caching, MCP server integration, safety hooks
• Setup Guide: .agent/CLAUDE_SETUP.md
• Settings Template: .agent/templates/claude.settings.json

Choose Claude if you use Claude Code IDE for best experience and advanced agent capabilities.

📊 DA Workflow Overview (Data Analytics Role)

The DA role comes with a complete multi-agent system for building data reports end-to-end:

Available Agents

| Agent | Purpose | AI Framework | |-------|---------|--------------| | orchestrator | Coordinates all 5 specialist agents | Opus 4.6 / inherit | | project-planner | Breaks down report tasks into phases | Sonnet 4.6 / inherit | | data-analyst | Metrics design, KPI definition, SQL analysis | Sonnet 4.6 / inherit | | data-engineer | ETL/pipeline design, ClickHouse optimization | Opus 4.6 / inherit | | database-optimizer | Query tuning, indexing, performance | Opus 4.6 / inherit | | backend-specialist | FastAPI endpoints, Pydantic models | Sonnet 4.6 / inherit | | frontend-specialist | React dashboards, D3.js visualizations | Opus 4.6 / inherit |

Getting Started with DA Workflow

# 1. Initialize DA role
npx @uocnv1998/agent-kit --role=da
# Choose: Antigravity or Claude

# 2. If Claude, copy settings template
cp .agent/templates/claude.settings.json ./settings.json

# 3. Read architecture & rules
cat .agent/ARCHITECTURE.md
cat .agent/rules/GEMINI.md  # or CLAUDE.md if Claude chosen

# 4. Start building reports!
# Invoke orchestrator for end-to-end report builds
# Or individual agents for specific tasks

Workflow Phases

1. ANALYSIS (data-analyst) - Define metrics and KPIs
2. DATA DESIGN (data-engineer) - Create SQL schemas and pipelines
3. OPTIMIZATION (database-optimizer) - Tune query performance
4. API (backend-specialist) - Build data APIs
5. UI (frontend-specialist) - Create interactive dashboards

🤝 Contributing

This project is built to improve the "Local-First, Handoff via Jira" strategy for AI Agent collaboration. If you have suggestions for new skills, role templates, or AI framework improvements, feel free to contribute!