All products

Review and refine interfaces with AI

Find a screen’s important problems and make focused changes within its existing design system

Review and refine interfaces with AI by starting with the screen’s purpose. The keyline skill examines interface code and rendered pages, ranks problems with evidence and separates that review from making changes. Your existing brand, token names and recorded decisions remain the basis for the work.

A project design file keeps role mappings, choices and open issues available for later sessions. Source checks can find some defects, while actual contrast, responsive presentation and working controls need browser inspection. Reports should distinguish measurements from design judgement and say what could not be tested. Internal scores are not accessibility certification.

What it does

  • Review rendered text contrast
  • Improve spacing and grouping
  • Align typography across a screen
  • Rank findings with source evidence
  • Record decisions for later sessions
  • Check narrow layouts and reduced motion

Install and use

  1. Prepare the complete skill in a compatible file/shell-capable agent. Use Node 18+, a writable project and a running application or preview for visual checks.
  2. For example: “Review the settings page at desktop and phone widths. Keep our brand colour and show the three most important problems before changing code.”
  3. Review the findings and choose a focused refinement. Have the agent map existing tokens and record decisions without introducing a competing naming system.
  4. Compare the result with the starting screen. Use the agent’s own browser or the maker’s browser companion, which may download Playwright and Chromium. Approve first-time setup and retain any untested limitations.

FAQ

Source

Made by Mihir Modi. Original skill: keyline, published on GitHub under the MIT licence.

View on GitHub