🚀 Gemini-Kit: 19 AI Agents Extension for Gemini CLI

[nth5693]

🚀 Gemini-Kit: Transform Gemini CLI into an AI Engineering Team

Hey everyone! 👋

I'm excited to share Gemini-Kit, an extension for Gemini CLI that adds 19 specialized AI agents and 15 skill modules to supercharge your development workflow.

✨ What is Gemini-Kit?

Gemini-Kit turns your terminal into a virtual engineering team with:

• 19 AI Agents: Planner, Coder, Tester, Reviewer, Debugger, Security Auditor, Frontend Specialist, Backend Specialist, DevOps Engineer, and more!
• 15 Skill Modules: React, Next.js, Docker, Security, API Design, Mobile, Performance, Tailwind, and more
• 44 Commands: /explore, /plan, /work, /review, /compound, /cook, /debug, /test...
• Compound Loop: Each task builds knowledge for faster future work
• Learning System: AI learns from your feedback
• Security: 30+ secret detection patterns

🔗 Links

• GitHub: https://github.com/nth5693/gemini-kit
• Quick Start: Clone, install, and start using in 2 minutes!

📦 Installation

git clone https://github.com/nth5693/gemini-kit.git ~/.gemini/extensions/gemini-kit
cd ~/.gemini/extensions/gemini-kit
npm install && npm run build
gemini extensions link $(pwd)

💡 Example Usage

# Research before implementing
/explore JWT authentication best practices

# Create detailed plan
/plan Add user authentication with JWT

# Execute the plan
/work

# Code review
/review src/api/auth.ts

🤝 Looking for Feedback!

I would love to hear your thoughts:

• What features would you find useful?
• Any issues or suggestions?
• Would you use something like this?

Feel free to try it out and let me know! PRs and issues are welcome on the repo.

Thanks! 🙏

[mystilleef]

Is there publicly available documentation for creating custom sub-agents. I tried writing custom agents on yesterday's nightly, but to no avail. Is there a more involved process than simply enabling agents in settings.json and dropping markdown directives in the ~/.gemini/agents folder?

[nth5693]

Thanks for trying out Gemini-Kit! 🙏

Great question! Currently, Gemini CLI's agent system is still evolving. Here's what I've learned from building Gemini-Kit:

How Agents Work in Gemini-Kit:

1. Agent files are markdown files in the agents/ folder with specific formatting
2. The key is the GEMINI.md file - it acts as the orchestrator that references all agents
3. Agents are referenced via the main instruction file, not loaded automatically from a folder

Quick Start:

# Clone and install
git clone https://github.com/nth5693/gemini-kit.git ~/.gemini/extensions/gemini-kit

# The agents are loaded through GEMINI.md which you can customize

Tips:

• Agents are essentially prompt engineering - they're markdown files with role definitions
• The GEMINI.md file in your project root acts as the "brain" that coordinates agents
• You can invoke agents by mentioning their role in your prompts

I'd be happy to help troubleshoot your setup! What specific agent behavior are you trying to achieve?