RouteMux Agent Skills

A few SKILL.md files that teach your coding agent to generate images and video and look up models with your RouteMux key. MIT-licensed, and they work in any agent that reads the open format.

Why you need them

Point a coding agent at any third-party gateway and its built-in image tool stops working. Codex CLI's $imagegen is the usual example: it rides the vendor's own subscription session and ignores the base_url you configured. This is not a misconfiguration and there is nothing to fix in your config. The fix is to give the agent a skill that calls the image endpoint over plain HTTP — that path is unaffected.

Four skills

routemux-images

Uses GPT Image 2.5 Flare by default. It turns your one-liner into a full prompt before drawing, and when you ask for a change it edits the last image instead of starting over.

routemux-video

Text-to-video with MiniMax H3 by default. It quotes the price first, then submits, waits and downloads the MP4 once you confirm.

routemux-models

Which models your key can call, what they cost, which capabilities and protocols they support.

routemux-setup

Point Codex CLI, Claude Code, Cursor, Cline and other clients at RouteMux.

Install

Any agent (detects Claude Code, Codex, Cursor and others on your machine)

npx skills add https://routemux.com -g

Claude Code plugin (type these in a session; needs 2.1.224 or later)

/plugin marketplace add https://routemux.com/.well-known/agent-skills/marketplace.json
/plugin install routemux@routemux

Install script (no Node needed, and it sets up your key)

curl -fsSL https://routemux.com/.well-known/agent-skills/install.sh | sh

Set your key

If Claude Code, Codex, Qwen Code or Gemini CLI already talks to RouteMux, the skills use that key and you can skip this. Otherwise set one environment variable: the install script can verify and save it for you; to set it by hand, add this line to ~/.zshrc or ~/.bashrc and open a new terminal.

export ROUTEMUX_API_KEY='sk-...'

Create an API key

Then just ask

The agent loads routemux-images, expands your sentence into a full image prompt, draws it with GPT Image 2.5 Flare and saves the file.

draw a paper map of the world folded into a single hop, save it to ./hop.png

What they touch, and what they do not

The key comes from ROUTEMUX_API_KEY, or is the key your agent already sends to api.routemux.com — taken from its environment variables, or from the RouteMux entry in the Codex or Claude Code config file. A key meant for anywhere else is never touched. They do not read .env, or any other credential or config file.
The skills never write the key into a file, a log or a commit. The only thing they write to disk is the file you asked for, at the path you gave. The install script saves the key to your shell profile only if you say yes.
No eval, no SDK install — just curl and a little python3. Every script is short enough to read in one sitting, and we would rather you did.

Read the integration docs