Guide · Agent skill
How to use the Generate Image agent skill: setup, prompts and tips
Updated · written with AI from the project's own documentation and checked against our catalogue data
The generate-image skill lets an AI agent create images, such as textures, icons, sprites, artwork and mockups, by sending your description to OpenAI or Google Gemini and saving the result into your project folder. This guide explains what an agent skill is, how to install this one, how to ask for images in plain language, and what you need to set up first, including API keys and the privacy trade-offs. Once you finish, you should be able to install the skill and generate your first image.
Generate Imagegithub/awesome-copilot · MIT Download the skill Quick start
- 1Download the ZIP. Get the generate-image skill ZIP from this page and unzip it to reveal the generate-image folder containing SKILL.md.
- 2Place the skill folder. Copy the folder to ~/.claude/skills/ for all projects, or to .claude/skills/ inside one project. You can alternatively upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Get an API key. Create an OpenAI key at platform.openai.com/api-keys or a Gemini key at aistudio.google.com/apikey.
- 4Set the key variable. Set SKILL_IMAGE_GEN_OPENAI_KEY or SKILL_IMAGE_GEN_GEMINI_KEY in your environment. If none is set, the agent will guide you through this in chat.
- 5Ask for an image. Describe the image, its purpose and any size or style needs in a normal request, such as a seamless texture or an icon.
- 6Check the saved file. The agent tells you where the image was saved. Open it and review it before you use it.
What is an agent skill, and what does generate-image do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a task matches the skill's description. The generate-image skill is one such folder. Its description tells the agent to use it when you ask to generate, create or make images, textures, icons, sprites, artwork, visual assets or mockups.
When the skill runs, the agent checks whether an API key for OpenAI or Google Gemini is set in your environment. It then sends your description to that provider's image API, receives the picture, and saves it to a sensible place in your project, such as assets/ or images/, or the current directory. It then tells you where the file went.
Install the generate-image skill
Download the skill ZIP from this page and unzip it. The skill is a single file, SKILL.md. Put the generate-image folder in one of these locations:
For all your projects, place it in ~/.claude/skills/. For one project only, place it in that project's .claude/skills/ folder.
You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/generate-image/
- Project install: .claude/skills/generate-image/ inside your project
Set up an API key
The skill needs an API key from at least one provider. If you already have a key set as SKILL_IMAGE_GEN_OPENAI_KEY or SKILL_IMAGE_GEN_GEMINI_KEY, the agent uses it. If neither is set, the agent walks you through the setup in conversation: it asks which provider you want, points you to the key page, and then saves the key you give it to your environment and shell profile.
OpenAI (gpt-image-2) is described as high quality with excellent text rendering, and it is paid per image. Google Gemini (Nano Banana) is described as fast, with a free tier available, and good for iteration. The key pages named in the skill are platform.openai.com/api-keys for OpenAI and aistudio.google.com/apikey for Gemini.
- OpenAI variable: SKILL_IMAGE_GEN_OPENAI_KEY
- Gemini variable: SKILL_IMAGE_GEN_GEMINI_KEY
- If both keys are set, the agent picks based on context or asks which you prefer.
Ask for an image with a natural request
You do not need a special command. Describe what you want and say what it is for. The skill's wording is built around requests to generate, create or make visual assets, so phrases like the ones below should match. For game work, the skill adds words such as seamless, tileable and game asset to the prompt on its own.
If you want several images at once, ask for a batch. The skill's guidelines say the agent makes multiple API calls in parallel for batch generation.
- Create a seamless stone floor texture for a dungeon game.
- Make a 32x32 pixel-style sprite of a red potion bottle.
- Generate three icon ideas for a note-taking app, flat style.
- Create a mockup illustration of a coffee shop landing page hero image.
- Make a batch of five forest background tiles for my platformer.
Settings the skill can use
The skill's documented options control the output. For OpenAI, you can choose the model (gpt-image-2 or gpt-image-1), the size (1024x1024, 1024x1536, 1536x1024 or auto) and the quality (low, medium or high). The defaults are gpt-image-2, 1024x1024 and medium.
For Gemini, the model is set in the request URL. The options listed are gemini-2.0-flash-exp (the default) and gemini-2.5-flash-image. You can ask the agent for a provider or option in your request, or ask what options are available and it will explain both providers.
- OpenAI size options: 1024x1024, 1024x1536, 1536x1024, auto
- OpenAI quality options: low, medium, high
- Gemini models: gemini-2.0-flash-exp, gemini-2.5-flash-image
Safety, privacy and cost notes
This skill does not run scripts. It does use the network: your prompts and your API key are sent to api.openai.com or generativelanguage.googleapis.com. If your project is sensitive and prompts should not leave your machine, do not use this skill with those providers.
The skill does not address image rights or review generated content. If you need licensed or brand-safe output, check each image yourself. OpenAI images are paid per image, and requests fail without a valid key. The skill also reports safety blocks from Gemini, which can occur when a prompt or output is filtered.
- Runs scripts: no. The agent makes HTTP requests and writes files.
- Network: yes. Prompts and keys go to third-party servers.
- Review every output before using it in a commercial product.
Tips for better results
Name the purpose and format in your prompt, such as whether the image is a texture, icon or mockup, and whether it should be transparent-background-free, pixel-style or flat. The skill adds game-asset cues on its own for game textures, so you can keep prompts short for those.
If a request fails, check that the right key variable is set, that the network can reach the provider's domain, and whether the reply mentioned a safety block. Ask the agent to try the other provider if one fails.
- Say where the image should go, for example assets/ or images/.
- Use size and quality options when a specific dimension matters.
- Start with Gemini for quick iterations, then use OpenAI for a final polish if you have budget.
What people use it for
- Placeholder textures and sprites for prototyping a 2D or 3D game.
- Icon concepts for an app or website during early design work.
- Mockup illustrations to show a client or team an idea quickly.
- Batch-generating a set of background or tile variations in parallel.
- Creating visual assets for documentation without leaving the agent chat.
Frequently asked questions
Is the generate-image skill free?
The skill itself is MIT licensed, according to its listing. Image generation depends on your provider: the skill says OpenAI images are paid per image, while Google Gemini has a free tier available. Check each provider's current pricing before you rely on it.
How do I install the generate-image skill?
Download the ZIP from this page and unzip it. Place the generate-image folder in ~/.claude/skills/ for all projects or in a project's .claude/skills/ folder for one project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
What do I need to use it?
You need an OpenAI API key (SKILL_IMAGE_GEN_OPENAI_KEY) or a Google Gemini API key (SKILL_IMAGE_GEN_GEMINI_KEY), network access to api.openai.com or generativelanguage.googleapis.com, and an agent that can make HTTP requests and write files.
Why does image generation fail?
The most likely causes are a missing or invalid API key, no network access to the provider domain, or a safety block. The skill documents Gemini responses that signal an error, a safety block or a filtered result, so read the reply for those details and try again or switch providers.
Is my data safe with this skill?
Your prompts and API key are sent to third-party servers at OpenAI or Google. Do not use it for sensitive projects where prompts must stay on your machine. The skill also does not review generated content or address image licensing.
How does it compare with other image tools?
This skill is for working inside an agent workflow with API-based generation from OpenAI or Gemini. If you need guaranteed licensing, brand-safe output, or a fully offline tool, pick a different option. The skill's own notes describe OpenAI as higher quality with paid usage and Gemini as faster with a free tier.
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
Use this skill if you are a developer or designer with an OpenAI or Google AI Studio account who wants placeholder assets, icon ideas or mockup images inside an agent chat. Pick something else if you have no API key or budget, need licensed or brand-safe images, or cannot send prompts to third-party servers.
Download the skill