Guide · Agent skill
How to use the Art image and diagram generator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Art is an agent skill that makes static visual content, such as blog header images, technical diagrams, framework charts, comparisons, comics, icons and wallpapers, using image models. It also turns some technical diagrams and essay illustrations into looping GIFs. This guide explains what the skill does, how to install it, how to trigger it with normal requests, and what it needs before you rely on it.
Art Image and Diagram Generatordanielmiessler/LifeOS · MIT Download the skill Quick start
- 1Download the ZIP. Download the Art skill ZIP from the page where you found this guide.
- 2Unzip into a skills folder. Extract the Art folder into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ for one project.
- 3Or upload it in the apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Add your image model key. Provide the API key for Nano Banana Pro, Nano Banana or Flux, following Setup.md.
- 5Check the runtime and tools. Make sure Bun or Node and the file utility are available on your machine.
- 6Run a request. Ask for a specific visual, such as a blog header or flowchart, and name its subject and use.
- 7Check the output. Open the image yourself to confirm labels, numbers and transparency are right before you publish it.
What the Art skill makes the agent do
An agent skill is a folder of instructions and tools that an AI agent loads when a request matches its description. Art covers static visual content in more than 20 formats, using Nano Banana Pro by default, Nano Banana, or Flux.
Each request is routed to a named intent, such as Essay, Mermaid, Frameworks, Comparisons or Comics. The intent encodes a technique and palette, then calls a bundled Generate.ts script that uses the image model. Generate.ts also checks that the intent was followed.
- Blog header images and essay illustrations
- Technical and architecture diagrams, static or animated as looping GIFs
- Mermaid flowcharts and sequence diagrams
- Frameworks, 2x2 matrices, comparisons, comics, icons and wallpapers
- Background removal and logo wallpapers
Install the Art skill
Download the Art skill as a ZIP from the page where you found this guide, then unzip it. Put the Art folder in one of two places: ~/.claude/skills/ for all your projects, or .claude/skills/ inside a single project.
If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there. Start a new conversation afterwards so the agent can see it.
Before the first real run, read Setup.md in the skill folder. It covers the image-model keys, rembg and ImageMagick, and a dry-run check.
- Option A: unzip into ~/.claude/skills/ (available everywhere)
- Option B: unzip into a project's .claude/skills/ (that project only)
- Option C: upload in the Claude apps under Settings, Capabilities, Skills
Set up image model access
Art needs access to an image model. The skill names Nano Banana Pro (Gemini 3 Pro) by default, plus Nano Banana and Flux. You supply the API key for the model you use, and Generate.ts references it.
The skill also needs a Bun or Node runtime to run its bundled TypeScript tools, and the file command-line utility, which verifies RGBA transparency. Output goes to your Downloads folder unless you set LIFEOS_DOWNLOADS_DIR to a different location.
- Image model API access and its API key
- Bun or Node runtime
- The file command-line utility
- Optional: LIFEOS_DOWNLOADS_DIR to change the output location
Trigger the skill with natural requests
You do not need special syntax. Describe the visual you want in plain words, and include a term from the skill's trigger list where you can. Terms the skill responds to include art, illustration, diagram, flowchart, animated diagram, diagram GIF, animated essay image, header image, blog social thumbnail, infographic, mermaid, comic, icon, framework diagram, D3 chart, remove background and wallpaper.
Be specific about the subject and the use. A request that names the article, the audience and the format gives the intent enough to choose the right technique.
- Make a header image for my blog post about remote work.
- Draw a flowchart of our order-processing pipeline.
- Create an animated diagram GIF of this system architecture.
- Turn this framework into a 2x2 matrix image.
- Make a comic panel series explaining how caching works.
- Remove the background from this logo and make a wallpaper.
Know what the skill will not handle
Requests that say visualize belong to a separate Visualize skill, which owns every 'visualize X' request and may embed Art output. YouTube and video thumbnails belong to the video skill. Other video or animation work belongs to HyperFrames, and web UI design belongs to Webdesign.
If you send these requests to Art, the output may be off target. Name the right skill or rephrase the request so it matches an Art trigger.
- 'Visualize' requests: Visualize skill
- YouTube or video thumbnails: video skill
- Other video or animation: HyperFrames
- Web UI design: Webdesign
Safety notes: scripts and network use
Art runs scripts and uses the network. Generation sends your prompts and any images you supply to external image-model services through their APIs. If you cannot send prompts or images to those services, do not use this skill for that work.
Generate.ts refuses to run unless it receives an --intent=<name> flag, or the explicit --freeform-confirmed opt-out. Review the intent and output path before you let it run on sensitive material.
- Prompts and reference images leave your machine for the image-model API
- Generation fails without an --intent flag unless you opt out explicitly
- Output defaults to your Downloads folder unless --output names another path
What people use it for
- Blog header images and social thumbnails that share one palette across a series of posts
- Architecture or flowchart diagrams for documentation, presented as static images or looping GIFs
- Framework, comparison and 2x2 matrix graphics for presentations and essays
- Comic panels and icons for explainers
- Logo wallpapers and background removal for brand assets
Frequently asked questions
How do I install the Art agent skill?
Download the ZIP from the page where you found this guide and unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings, Capabilities, Skills.
Is the Art skill free?
The skill is released under the MIT licence. The source does not state prices for the image-model APIs it calls, so check your provider's pricing before you generate large batches.
What do I need to run the Art skill?
You need API access and a key for an image model (Nano Banana Pro, Nano Banana or Flux), a Bun or Node runtime, and the file command-line utility. Setup.md in the skill folder covers the keys and dependencies.
Why does the skill ignore my request to visualize data?
Requests that say visualize are owned by the separate Visualize skill, which may embed Art output. Use the Visualize skill for those requests, or rephrase so the request matches an Art trigger such as diagram or infographic.
Can Art make YouTube thumbnails?
No. YouTube and video thumbnails are handled by the video skill, which has its own art intents and compositors. Art may produce off-target output for those requests.
Where are the generated images saved?
Output goes to your Downloads folder by default. Set the LIFEOS_DOWNLOADS_DIR environment variable to change that location, or pass a different output path to Generate.ts.
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
Art suits bloggers who need consistent header images, people making polished technical diagrams, and designers who want a shared style across a set of images. It is a poor fit if you cannot send prompts and images to external image-model APIs, or if your request is a generic visualization, a video thumbnail or web UI design, where another skill is the better choice.
Download the skill