Guide · Agent skill
How to use the Gemini CLI Assistant agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The gemini agent skill teaches an AI agent to call Google's Gemini command-line tool for single-question tasks such as answering a question, summarising text, or generating a draft. This guide is for people who may never have used agent skills before, and it walks through what the skill does, how to install it, how to ask for it in plain language, and what to watch for. By the end you will know how to set it up and how to get structured output from Gemini inside an agent workflow.
Gemini CLI Assistantspinabot/brigade · MIT Download the skill Quick start
- 1Install the Gemini CLI. Install the gemini-cli Homebrew formula so the gemini binary is available in your terminal.
- 2Run gemini once. Run gemini interactively and follow the login flow with your Google account or Gemini API credentials.
- 3Download and unzip the skill. Download the ZIP from this page and unzip it to get the gemini folder containing SKILL.md.
- 4Place the skill folder. Move it to ~/.claude/skills/ for personal use or to a project's .claude/skills/ folder.
- 5Ask for a one-shot answer. Make a natural request such as asking Gemini to summarise a paragraph, and the agent will run a one-shot prompt.
- 6Request JSON or a model. Mention the output format or model you want, and the agent can add the matching flag.
What the Gemini skill makes the agent do
An agent skill is a folder of instructions that an AI agent reads when a task matches the skill. This one is a single file, SKILL.md, and it contains only instructions. It does not add new code of its own.
With the skill loaded, the agent sends a prompt to Gemini from the terminal in one-shot mode, meaning one question and one answer rather than a chat session. It can choose a specific model and can ask for JSON output. It also knows how to list and manage Gemini CLI extensions.
- Answer a single question through Gemini
- Summarise text or generate content
- Choose a specific model for a prompt
- Return JSON for use in a script
- List and manage Gemini CLI extensions
Install the Gemini CLI before you install the skill
The skill needs the Gemini CLI, which provides a binary called gemini. The skill's metadata lists a Homebrew install option using the formula gemini-cli. Install the CLI first, then add the skill.
After installing, run gemini once interactively and follow the login flow. Authentication uses a Google account login or Gemini API credentials. The skill does not explain how to set up billing or API keys, so check Google's documentation at https://ai.google.dev/ for that.
brew install gemini-cliInstall the skill from the ZIP
Download the ZIP from this page and unzip it. For a personal install, place the skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills
How to trigger the skill with natural requests
You do not need a special command. Ask the agent for a quick answer, a summary, or a generated draft, and say that Gemini should handle it. The agent uses the skill when the request matches its description: Gemini CLI for one-shot Q&A, summaries, and generation.
If you want a particular model or JSON output, say so in the request and the agent can pass those options to the gemini command.
- "Use Gemini to summarise this paragraph in three sentences."
- "Ask Gemini which of these two error messages is more likely to be a permissions problem."
- "Have Gemini return the list of action items as JSON."
- "Use the Gemini model named in the flag to draft a short product description."
Command reference the agent uses
These are the commands the skill documents. The agent runs them in one-shot mode with a positional prompt, which means the prompt is passed directly after the command.
To list extensions, the agent uses the list flag. To manage them, it uses the extensions subcommand with a command you supply.
- gemini "Answer this question..." for a basic one-shot prompt
- gemini --model <name> "Prompt..." to pick a model
- gemini --output-format json "Return JSON" for structured output
- gemini --list-extensions to list extensions
- gemini extensions <command> to manage extensions
gemini "Answer this question..."Safety notes: scripts, network and --yolo
This skill runs scripts and uses the network. Every prompt is sent to Google's Gemini service, so do not use it with sensitive or confidential documents.
The skill tells the agent to avoid interactive mode, so long chat sessions are not its purpose. It also tells the agent to avoid the --yolo flag for safety. Keep that rule in place and do not ask the agent to use that flag.
- Prompts travel over the network to Google's Gemini service
- Do not send confidential or sensitive documents
- Avoid interactive mode for long chats
- Avoid the --yolo flag
Tips for better results
Keep prompts self-contained, because each call is a single question with a single answer. If you need structured data for a script, ask for JSON output explicitly and check the result before using it.
If a call fails with an authentication message, run gemini once interactively to complete the login flow, then retry.
- Put all the context the model needs into one prompt
- Ask for JSON when a script will parse the output
- Re-run gemini interactively if auth is required
What people use it for
- Summarise a short article or meeting note into a few sentences
- Generate a first draft of a product description or email
- Answer a single factual or reasoning question during an agent task
- Get structured JSON output from Gemini for an automated script
- List installed Gemini CLI extensions and manage them from the agent
Frequently asked questions
How do I install the gemini agent skill?
Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
Do I need the Gemini CLI installed first?
Yes. The skill calls the gemini binary, which the skill's metadata says comes from the Homebrew formula gemini-cli. Install that and run gemini once to log in before using the skill.
Is the Gemini CLI free to use?
The source does not state pricing or licence terms for the Gemini CLI or the Gemini API. The skill itself is MIT licensed. Check Google's documentation at https://ai.google.dev/ for current access details.
Why does the agent say authentication is required?
The tool cannot run without authentication. Run gemini once interactively and follow the login flow with your Google account or Gemini API credentials, then try again.
Can I use this skill for confidential documents?
Not safely. Prompts are sent to Google's Gemini service over the network, so avoid sensitive or confidential material.
What are alternatives if Gemini is not suitable?
The source does not compare this skill with other tools. If you need long interactive chat sessions or a fully local model, look for a different skill or tool, since this one tells the agent to avoid interactive mode.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits developers who already have the Gemini CLI set up and want their agent to make quick one-shot calls for answers, summaries, drafts, or JSON output. It is a poor fit if you have no Gemini access, must keep documents confidential, or want long interactive chat.
Download the skill