Gemini CLI Assistants

A collection of ready-to-use Gemini CLI agent templates for business workflows. Quickly bootstrap new projects with best-practice agent settings and documentation.

1. Prerequisites

• Node.js (v16+ recommended)
• npm or yarn

Check your installation:

node -v
npm -v

2. Install Gemini CLI

Before installing the Gemini CLI, ensure you have Node.js (v16+ recommended) installed on your system. You can install it via your system's package manager or directly from the Node.js website.

For macOS and Linux

Install the Gemini CLI globally using npm or yarn:

npm install -g @google/gemini-cli
# Or with yarn:
yarn global add @google/gemini-cli

For Windows

1. Install Node.js (if not already installed): Using Chocolatey (recommended package manager for Windows):

   choco install nodejs-lts -y

   Alternatively, download the installer from the Node.js website.

2. Install Gemini CLI: Open PowerShell or Command Prompt as an administrator and run:

   elevate npm install -g @google/gemini-cli

   (The elevate command is often available if you have Chocolatey installed, or you can manually run your terminal as administrator.)

3. Get Your Gemini API Key

To use the Gemini CLI, you need a Gemini API key.

1. Go to Google AI Studio and sign in with your Google account.
2. Create or view your Gemini API key.
3. Set the API key as an environment variable in your shell (preferred method):

   export GEMINI_API_KEY=your-api-key-here

   Or follow the CLI's instructions for configuring the API key if different.

Keep your API key secure and do not share it publicly.

4. Project Setup

Follow these steps to create a new project using a Gemini CLI assistant template:

1. Clone this repository somewhere outside your target project:

   git clone https://github.com/devkindhq/gemini-cli-assistants
   cd gemini-cli-assistants

   Alternative: Download Manually (No Git) If you don't have Git installed or prefer not to use it, you can download the repository as a ZIP file:

   1. Go to the Gemini CLI Assistants GitHub page.
   2. Click the green <> Code button.
   3. Select Download ZIP.
   4. Extract the contents of the ZIP file to a location on your computer (e.g., ~/Downloads/gemini-cli-assistants-main).
   5. Navigate into the extracted directory:

      cd ~/Downloads/gemini-cli-assistants-main

   Then, proceed with the rest of the setup steps, adjusting paths as necessary.

2. Create a folder for your Gemini CLI projects (if you don't have one):

   mkdir -p ../gemini-cli-projects
   cd ../gemini-cli-projects

3. Create a new project folder (e.g., shopify-redesign-proposal):

   mkdir shopify-redesign-proposal
   cd shopify-redesign-proposal

4. Use an agent template

   • The agent templates are already located in their respective folders under assistants/.
   • To use an agent, simply navigate to its folder (e.g., assistants/sales-strategy/) and run the gemini command from there.
   • If you modify an agent's GEMINI.md or .gemini/settings.json file, you may need to close and reopen the Gemini CLI for the changes to take effect.
   • projects/: A folder where users can store project-specific information, documents, or data that the agents can reference. This folder is ignored by Git.

5. Project Structure Example

After navigating into an agent's directory (e.g., assistants/sales-strategy/), your project structure will look like this:

assistants/
└── sales-strategy/
    ├── GEMINI.md
    └── .gemini/
        └── settings.json

• GEMINI.md: The agent's system prompt and documentation.
• .gemini/settings.json: The configuration file for the Gemini CLI assistant, including tool definitions.

Usage

Once set up, run the Gemini CLI in your project directory to interact with your chosen assistant:

gemini

Tips & Shortcuts

• Press the Up Arrow in the terminal to quickly access your previous Gemini CLI conversations.
• You can ask Gemini to write information directly into files, so you don’t have to copy-paste content manually.
• Type /mcp in the Gemini CLI to check which MCPs (Multi-Command Plugins) are available.
• Press Ctrl+T in the Gemini CLI to view details about each MCP—helpful for discovering what’s available and what’s working.

Caveats

If you have any environment variable that matches CI at the start of its name (e.g., CI_TOKEN, CI_BUILD), the Gemini CLI might interpret this as running in a Continuous Integration (CI) environment and disable its interactive prompt interface. To temporarily enable the prompt interface, unset such a variable for the specific command. For example:

env -u CI_BUILD gemini

To check if you have any CI related environment variables set, you can use:

env | grep ^CI

Available Agents

Here's a list of the pre-configured Gemini CLI agent templates available in this repository:

Sales Strategy Assistant

Description: A senior AI assistant designed to drive sales and identify new opportunities by proactively hunting for leads, analyzing strategic gaps, and delivering action-oriented outputs.

Read More

Technical Research Assistant

Description: A senior AI research analyst and strategic partner for the engineering team. Its purpose is to execute technical research with rigor, precision, and speed, transforming ad-hoc research into a disciplined, data-driven process.

Read More

Use Cases

Gemini CLI Assistants are versatile and can be used for a variety of collaboration and workflow scenarios:

• Engineering, DevOps, QA, UI/UX, and more:
  • Collaborate on technical projects, automate workflows, and streamline team communication.
  • Advanced: Initialize a git repository in your project folder to track changes, share knowledge, and collaborate with your team:

    git init
    git add .
    git commit -m "Initial commit"

• Marketing, Sales, Project Management, etc.:
  • For non-technical teams, simply place your project folder inside a shared drive (Google Drive, OneDrive, Dropbox, etc.).
  • Use the Gemini CLI from your desktop in that folder—no technical setup required. Desktop app versions of these drive services are available for easy access.

Model Context Protocols (MCPs)

For more information on Model Context Protocols and how they extend the capabilities of the Gemini CLI, refer to the MCP README.

Resources

• Gemini CLI Configuration Guide
• Get your Gemini API Key
• Model Context Protocols (MCPs)

Leaked and Open-Source System Prompts

Understanding how large language models are prompted is crucial for effective interaction and for developing robust AI applications. This section provides resources on leaked prompts and publicly available system prompts, offering insights into real-world prompting strategies.

• Zed.dev Leaked Prompts
• Collection of Leaked System Prompts (GitHub)
• A List of Leaked System Prompts (mattrickard.com)
• FULL v0, Cursor, Manus, Same.dev, Lovable, Devin, Replit Agent, Windsurf Agent, VSCode Agent, Dia Browser, Xcode, Trae AI & Cluely (And other Open Sourced) System Prompts, Tools & AI Models (GitHub)
• Collection of extracted System Prompts from popular chatbots like ChatGPT, Claude & Gemini (GitHub)

Contributing

Contributions are welcome! Please open issues or pull requests for improvements or new agent templates.

License

This project is licensed under the MIT License. See LICENSE for details.