Skip to content
Get packs

Guide · Agent skill

How to use the Image & Graphic Design Matrix agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

This guide explains what the Image & Graphic Design Matrix agent skill does, how to install it, and how to ask for image work in plain language. It is written for readers who have never used an agent skill before. By the end you will know what the skill needs on your machine, how it chooses between tools, and where it will stop and ask you first.

Image & Graphic Design MatrixHKUDS/CLI-Anything · Apache-2.0

Quick start

  1. 1Download the ZIP. Get the cli-hub-matrix-image-design skill ZIP from the page where you found this guide.
  2. 2Unzip into the skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
  3. 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the ZIP.
  4. 4Run the preflight check. Run cli-hub matrix preflight image-design --json to see which tools, packages and keys are available.
  5. 5Make a plain request. Describe the image job, the output format and whether it must stay offline.
  6. 6Approve any paid provider. When the agent asks which paid or metered provider to use, reply with your choice or skip it.

What the skill makes the agent do

The skill teaches an AI agent to choose among many image and design tools for jobs such as generating pictures, editing photos, making vector graphics, drawing diagrams and upscaling images. The tools it knows about include GIMP, Krita, Inkscape, draw.io, ComfyUI, and several paid AI image APIs.

The skill's main instruction is to pick a provider based on your requirements, output quality, whether you need offline work, which credentials are present, and install cost. It does not simply use the first tool in its list.

  • Raster editing: GIMP, Krita, Pillow, OpenCV, scikit-image, rembg for background removal
  • Vector editing: Inkscape, svgwrite, cairosvg, svgpathtools, vpype
  • AI image generation: ComfyUI, diffusers, and paid APIs such as OpenAI GPT-Image-1 and Stability AI
  • Diagrams: draw.io, Mermaid, Graphviz and PlantUML
  • Upscaling: Real-ESRGAN, GFPGAN, upscayl and paid options

Install the skill from the ZIP or the Claude apps

Download the skill ZIP from the page where you found this guide. For Claude Code, unzip it into your personal skills folder at ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill limited to one project.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the ZIP there.

  • Personal install: ~/.claude/skills/
  • Project install: .claude/skills/ inside your project folder
  • Claude apps: Settings → Capabilities → Skills, then upload

Check what is already installed on your machine

The skill begins by checking which tools and keys are present. You can run the same check yourself. The skill's own preflight command reports availability; it is not a provider selector.

The check looks for command-line tools such as inkscape, gimp, krita, drawio, magick and exiftool, and for Python packages such as Pillow and diffusers. It also reports whether API keys such as OPENAI_API_KEY and REPLICATE_API_TOKEN are set, without showing their values.

  • Requires the cli-hub command, used for the matrix install and preflight
  • Optional command-line tools, Python packages and API keys are only needed for the providers you plan to use
  • Offline work is possible with GIMP, Krita, Inkscape and local diffusers for most tasks
cli-hub matrix install image-design cli-hub matrix info image-design cli-hub matrix preflight image-design --json

Trigger the skill with a natural request

You do not need to type a command to use the skill. Describe the image job in plain words and mention the result you want. The agent should then check what is available and choose a suitable tool.

Be specific about the output type, whether the work must stay offline, and whether you are willing to pay for a service. If you do not state these, the agent will ask before using any metered or paid API, and it will ask which provider you want when several paid options have keys set.

  • "Make a vector icon of a padlock and export it as SVG."
  • "Draw a flowchart of our login process as a diagram I can edit."
  • "Upscale this photo to four times its size, offline if possible."
  • "Remove the background from product.jpg."
  • "Generate a 1200 by 630 social card; don't use any paid service."

Safety, network and cost notes

The skill runs scripts and uses the network. Some providers are cloud services, so files you send to them leave your machine and go to a third party. Do not use cloud providers for confidential or sensitive images unless you accept that.

The skill asks for consent before metered APIs, but a provider you have not vetted can still charge you. Review the provider's pricing before you approve it. The skill also lists Figma as a known gap, so it does not give you a working Figma path.

  • Files sent to cloud APIs leave your machine
  • Paid and metered providers require your consent
  • Figma is not supported by a working path in this skill

Tips for better results

Install the local tools you care about before you ask for work. A local GIMP, Krita or Inkscape install gives the agent options that need no key and no upload. Set only the API keys you intend to use, since the agent treats keys that are present as available credentials.

If a provider is not what you expected, say so and name the tool you want, for example Inkscape rather than a generative model. Use the preflight output to see what the agent can actually reach on your machine.

  • Name the output format (SVG, PNG, Mermaid, draw.io) in your request
  • Say "offline only" when you need no network use
  • Keep API keys to the providers you have chosen

What people use it for

  • Create a vector icon in Inkscape and export it as SVG for a website
  • Draw an editable flowchart or architecture diagram with draw.io or Graphviz
  • Remove a background from a product photo with rembg
  • Upscale a low-resolution image with Real-ESRGAN or upscayl
  • Generate a social sharing image with a local ComfyUI or diffusers setup

Frequently asked questions

How do I install the image design agent skill?

Download the skill ZIP from the page, then unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

Is the image design skill free?

The skill's own instructions are provided under the Apache-2.0 licence. Many of the tools it can use are free and local, such as GIMP, Krita, Inkscape and Pillow, but several AI image APIs and some other services are metered or paid, and the skill asks for consent before using them.

What does the skill require to work?

It requires the cli-hub command. The other tools are optional: command-line programs such as inkscape, gimp, krita, drawio and magick, Python packages such as Pillow and diffusers, and API keys for paid services. You only need the ones for the providers you plan to use.

Can I use the skill completely offline?

Yes for many tasks. The skill notes that GIMP, Krita, Inkscape and local diffusers cover almost everything offline. Cloud APIs and some services will need network access, so choose offline providers if that matters to you.

Does the skill support Figma?

No working Figma path is provided. The skill lists Figma as a known gap, even though a Figma API entry with a FIGMA_TOKEN appears in its mockup table.

Why does the agent ask me before using some tools?

The skill tells the agent to escalate to paid or metered APIs only when credentials are already present or you explicitly consent. When several credentialed AI image providers are available, it asks which one you want.

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 designers, developers and marketers who already have some local image or diagram tools installed and want an agent to pick sensible options and use them. It is a poor fit if you need Figma, want unattended paid generation, or must keep sensitive images off cloud services; in those cases use the tools directly or choose a different workflow.

Download the skill