Three free Agent Skills (MIT license)

Design and check for people and agents

Three free Agent Skills help you build AI-native pages, which are pages made for people and AI agents. Load them in Figma or coding harnesses, which are tools that run coding agents.

How they work

How the 3 skills work together

  1. 01

    Design

    Use basic-ux-figma or basic-ux-design so your agent builds for people and web agents.

  2. 02

    Build

    Your agent writes clean HTML code and a clear accessibility tree for AI agents.

  3. 03

    Audit

    Run basic-ux-audit in a browser to score the page from 0 to 100.

  4. 04

    Fix

    Fix each failed check and run the audit again until both scores reach 90.

The skills

What is inside each skill

Design skill

basic-ux-design

This skill tells a coding agent how to build a clear page for people and AI agents. It turns the 5 principles into rules.

  • SKILL.md gives the build rules for all 5 principles.
  • references/agent-ready.md gives code for labels, llms.txt, structured data and WebMCP tools.
  • references/visual-craft.md covers fonts, color, layout, motion and AI design habits to skip.
  • brand.example.json holds your brand colors and fonts.

Audit skill

basic-ux-audit

This skill opens a page in Playwright, which is a browser test tool, and scores the 5 principles. Then your agent checks the screen pictures.

  • SKILL.md gives the scoring rules and check steps.
  • scripts/assess.mjs checks the live page in a browser with Playwright.
  • report.md and report.json list scores and failed checks by principle.
  • agent-view.yaml shows the page as web agents read it.

Figma skill

basic-ux-figma

This skill checks Figma frames for people and writes notes for AI agents. Coding agents read those notes to build the page.

  • SKILL.md runs in Figma Make, Figma's agent or the Figma MCP server, which links Figma to agents.
  • People checks review color contrast, text size, tap size and spacing.
  • BASIC UX: agents annotations add button names, headings, landmarks and labels.
  • Score reports give 10 principle scores and top fixes.
$ node scripts/assess.mjs https://basicux.com

BASIC UX audit: basicux.com
Overall 97  (people 93, agents 100)

                 People  Agents
  Beauty           100     100
  Accessibility    100     100
  Simplicity        86     100
  Intuitiveness    100     100
  Consistency       78     100

  No measured failures. Finish the judgement steps in report.md.

Wrote report.md, report.json, agent-view.yaml and screenshots to ./basic-ux-report
Real output of the audit skill run against basicux.com

Why

Four reasons to check both together

Agents read the accessibility tree

AI agents read the browser's accessibility tree. AudioEye found in September 2026 that AI agents finished 96% of tasks on an accessible page.

Good pages cut failed steps

Elnaffar and Rashidi showed in July 2026 that a rebuilt store page raised agent success to 89.3%. Agents also took 30% fewer steps.

Bad labels break agent tasks

In September 2026, MERJ checked a button with a wrong hidden code label. An AI agent clicked Cancel instead of Save in 25 of 25 runs.

Separate AI pages hurt people

Adrian Roselli said in October 2025 that AI-only code can make pages worse for disabled people. Jakob Nielsen wrote in July 2026 to give people and agents 1 view.

Scores

Ten scores across 5 principles

Every run gives you 5 scores for people and 5 scores for AI agents, from 0 to 100.

Beauty

PeopleIs it aesthetically pleasing?

AgentsIs the page code clean?

Accessibility

PeopleCan everyone use it?

AgentsCan an agent read and use it?

Simplicity

PeopleDoes it make life easier?

AgentsDoes it take the fewest steps?

Intuitiveness

PeopleIs it easy to learn?

AgentsDoes it say what it does?

Consistency

PeopleDoes it match the system?

AgentsDoes it act the same every time?

BASIC UX audit works with Google Lighthouse's 2026 Agentic browsing category and Cloudflare's isitagentready.com. It adds the people checks and the 5 principles.

Where they run

Where the skills run

The Agent Skills install into Claude Code, Codex, Cursor and GitHub Copilot with npx skills add https://basicux.com.

Download

Get the 3 skills

Get all 3 skills in 1 zip file or get each skill alone. Use the form below to add your brand colors and fonts first.

Add your brand settings

Start

Checked on this page

Download all three skills Read the principles

This page scores 97 in basic-ux-audit, with 93 for people and 100 for AI agents.