Agent Skills

This repository includes an AI agent “skill pack” for rEFui at skills/refui. It is designed to help you implement requirements using rEFui’s existing primitives first (signals, built-in components, extras, renderers), and only fall back to custom code when rEFui doesn’t already provide a suitable solution.

Install

The skill lives at skills/refui in this repo.

For Codex CLI

Install it by copying it into your $CODEX_HOME/skills folder:

mkdir -p "$CODEX_HOME/skills"
cp -R skills/refui "$CODEX_HOME/skills/refui"

For Gemini CLI

Install the skill from the project root:

mkdir -p ./gemini/skills # for workspace
mkdir -p ~/.gemini/skills # globally
cp -R skills/refui ./gemini/skills

Activate the skill:

• "/skills enable refui" (in interactive mode)

For other agents

• In the installed folder, treat SKILL.md as the entrypoint “playbook”.
• Load references/*.md selectively for the task at hand (async UI, portals, lists/perf, etc.).
• Optionally run scripts/refui-audit.mjs from inside the installed skill folder as a lightweight project scan (Node.js required).

Use

Mention the skill name (or otherwise enable the pack in your agent) so it loads its workflow + references.

Examples (for Codex CLI):

• “Use the refui skill to implement a modal with a portal.”
• “Use $refui and help debug why my UI isn’t updating.”

Examples (for Gemini CLI):

• “Activate the refui skill and implement a counter.”
• “Use the refui skill to refactor this component to use signals.”
• “/skills enable refui” followed by “Help me optimize this list rendering.”

When you’re unsure

If you’re using rEFui as a dependency and can’t inspect its source, the skill instructs agents to consult MCP documentation sources (Context7 / DeepWiki) before guessing (when MCP is available).