Gemini Kit

Gemini Kit is a multi-agent AI development framework that runs inside the Gemini CLI. It orchestrates specialized AI agents and atomic skills to deliver high-quality software engineering outcomes with a focus on architectural integrity and token efficiency.

🚀 Quick Start

Prerequisites

• Gemini CLI installed.
• Node.js (v18+) and Python (v3.10+) available.

Installation

1. Install CLI (Recommended)

npm install -g geminicli-kit

This installs the gk command globally. You can then scaffold Gemini Kit into any project:

cd my-project
gk init      # Scaffold .gemini/ structure, agents, and skills
gk list      # List available agents and skills
gk update    # Pull the latest framework updates
gk uninstall # Remove Gemini Kit from the current project
gk bridge    # Manage Claude-Gemini bridge pipeline
gk version   # Show installed Gemini Kit version

2. Clone for Contribution

git clone https://github.com/hieund-it/geminikit.git
cd geminikit
npm install

For more details, see the Setup Guide.

📚 Core Documentation

All documentation is located in the docs/ directory:

• Onboarding Guide: The best place to start. A rapid introduction to the framework.
• System Architecture: Explains how the Orchestrator, Agents, and Skills work together.
• Agent Registry: A complete list of all available agents and their specializations.
• Commands Reference: A guide to using the /gk- commands to interact with agents.
• Skills Guide: A reference for all atomic skills available to agents.

🧩 Core Concepts

The framework is built around a few key ideas:

• Orchestrator: The central brain that parses commands and delegates tasks.
• Agents: Specialized, role-based AIs (e.g., Developer, Reviewer) that own specific domains.
• Skills: Atomic, reusable tools (e.g., gk-debug, gk-git) that agents use to execute tasks.

✨ Key Features

• 🧠 Advanced Memory System: Features auto-persistence, silent summarization, and self-healing state management to ensure seamless long-running sessions.
• 🛡️ Robust Security Framework: Implements a strict Agent Permission Matrix, forbidden path blacklisting, and pre-emptive secret masking for safe execution.
• 📉 Context Economy: Built-in token optimization and context rotation to minimize costs and maximize model performance.

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for details on how to submit pull requests, report issues, and propose new features.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.