Skip to the guide
All guides

Interface design

Design, build or review WordPress user journeys, responsive layouts, UI states and design tokens. Checks editor and front end parity, RTL and visual regressions.

app WordPress DevKit 2.0.3checked reading 4 minutes

On this page

When to use it

Use this skill when a page or flow needs better layout, states or consistency. It covers user journeys, responsive layouts, UI states, design tokens, parity between the editor and the front end, right-to-left (RTL) layout and visual regression checks.

It does not audit WCAG criteria, explain theme.json internals or build wp-admin screens.

Where it stops

  • WCAG criterion-by-criterion conformance: wp-devkit-wcag-review. Accessibility defects in code: wp-devkit-accessibility-review.
  • Theme templates and theme.json mechanics: wp-devkit-theme-development. wp-admin screens: wp-devkit-admin-ui-development. Block internals: wp-devkit-block-development.
  • Speed measurement and tuning: wp-devkit-performance-review.

Run a review in Claude

With the plugin loaded, start a read-only review of a bounded target:

Text
/wp-devkit:design-review <target-path>

The short form /wp-devkit:design asks for an implementation. Use the review form when you want the agent to stay read-only.

Run a review in Codex, ChatGPT or Antigravity

Name the installed skill and keep the request small. Replace the target with your files or journey.

Text
$wp-devkit-ui-design
Review the target without editing files.
Read the project instructions and the skill's engineering contract first.
Separate confirmed findings from candidates and state the evidence and checks you could not run.

The engineering contract is the set of rules every DevKit skill follows. It defines how findings are classified and how severe each one is.

Reference files

The agent loads only the file that matches your task.

  • wordpress-ui-surfaces.md: used to find which part of WordPress owns a screen or element.
  • design-systems-and-tokens.md: used for colors, spacing, type and other design tokens, and where they live.
  • responsive-states-and-interaction.md: used for layout at different widths, loading, empty, error and success states, and keyboard and touch behavior.
  • localization-and-content-design.md: used for translation, RTL and the wording of interface text.
  • research-and-usability.md: used when you make claims about usability and need evidence for them.
  • analytics-and-experiments.md: used when you make claims about conversion and need measurements or tests.
  • visual-regression-and-devices.md: used for the rendering matrix, Playwright baselines, Core Web Vitals evidence and axe automation.
  • acceptance-checklist.md: used to decide whether delivered work is accepted.
  • ui-design-workbook.md: used to triage symptoms, work through example decisions, recognize harmless look-alikes and choose verification steps.

How a review runs

  1. The agent captures the current journey and content: URLs, viewports and screenshots. It defines the outcome the user wants and separates visual polish from a structural rebuild.
  2. It finds the owner of each surface and token (wordpress-ui-surfaces.md, design-systems-and-tokens.md). Working navigation, links and content stay as they are.
  3. It checks states, forms, layout and interaction (responsive-states-and-interaction.md). That means loading, empty, error and success states, long content, widths from 320 px up, zoom, keyboard use, focus visibility, target size and reduced motion.
  4. It checks the locale and RTL layout with stress data such as long words and translated strings (localization-and-content-design.md).
  5. For any claim about usability, conversion or speed, it names the evidence available (research-and-usability.md, analytics-and-experiments.md). A claim without evidence is marked "not measured".
  6. It reports the rendered and interaction evidence it observed, plus proposed acceptance checks. A review does not authorize a redesign.

The agent reads ui-design-workbook.md for the decision table, symptoms, false positives, severity and a worked example.

What you get back

A review starts with the result and what was observed: viewports, browsers, locales and any assistive technology actually used. Each confirmed concern gives its location (file and line, or URL and viewport), who is affected, the trigger, the path that reaches it, the impact, the confidence, a minimal fix and a regression check. Candidates and unmeasured claims are listed apart.

An implementation report covers the user flow, responsive and keyboard checks, screenshots when available, the empty, loading, error and success states, commands with exit codes and the checks that were not run. A visual pass does not prove usability, conversion, screen-reader behavior or conformance.

Note

Reviews are read-only. A clean review covers only the inspected scope and does not certify the project. Ask for implementation as a separate, bounded task with a regression check.

Where to go next

Use review to fix to turn findings into a bounded change. Use Doctor and reports for environment and evidence tooling.