agentleFS
vishal-mandhanevishal-mandhane/divinity/skills/ui-ux-review/SKILL.md

ui-ux-review

Skill3 starsChanged 50 days agoMITView on GitHub

Use when reviewing or changing any app UI for accessibility, responsiveness, dark mode, motion or loading states — screen reader labels, touch target sizes, contrast ratios, text scaling, overflow on small screens, reduced motion, and missing empty/error states. Invoke after building or editing a screen or component, before shipping a UI change, or when asked "is this accessible", "does this work in dark mode", "check for overflow", or "review this screen". Examples are Flutter; the checks apply to any stack.

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. adtn0810/truestack★ 2same name
    2. dafaalz/agent-skills★ 2same name
    3. Kaidera-AI/skills★ 2same name
    4. zehuichan/vibe-harness★ 2same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository4 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 7 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Checked or unchecked, never assumed: If you did not render it at 2× text or on a 320-wide screen, say "not rendered" — do not pass it.

    line 24
  2. Distinctive:

    The project's design system wins on style: If the repo has a theme file, design tokens, or a design-system doc, hardcoded values that …

    line 26
  3. Distinctive:

    Test these two conditions, or mark them "not rendered":

    line 61
  4. Very specific:

    Every colour comes from the theme (Theme.of(context).colorScheme, or the project's tokens).

    line 96
  5. Distinctive:

    Check both modes for contrast.

    line 99
  6. Very specific:

    Avoid pure black #000000 surfaces behind pure white text for long reading — the halo is tiring.

    line 101
  7. Distinctive:

    No flashing over 3 per second (see §1).

    line 119
  8. Distinctive:

    Every screen that loads data needs all four.

    line 123
  9. Distinctive:

    No network, and a slow network (does loading ever end?)

    line 147

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

UI/UX Review

Overview

If a user can't see it, reach it, or understand it, it isn't there.

This skill checks one screen or component against measurable rules. Every finding names a number from a published standard, the file:line that breaks it, and who is hurt by it. "Looks a bit cramped" is not a finding.

Read the whole file first. A screen's error state is usually 300 lines below its happy path.

The Iron Laws

  1. Cite the standard, not taste. Each rule below has its source. If a finding has no number and no source, it goes under "Polish", not "Issues".
  2. Every finding has file:line from a file you opened this session.
  3. Checked or unchecked, never assumed. If you did not render it at 2× text or on a 320-wide screen, say "not rendered" — do not pass it.
  4. The project's design system wins on style. If the repo has a theme file, design tokens, or a design-system doc, hardcoded values that bypass it are findings. This skill never overrides the project's own colours or fonts.

1. Accessibility (the numbers)

CheckRuleSource
Text contrast4.5:1 for normal text, 3:1 for large text (≥ 18pt, or ≥ 14pt bold)WCAG 2.2 SC 1.4.3 (AA)
Non-text contrast3:1 for icons, input borders, focus rings, and state indicators against their backgroundWCAG 2.2 SC 1.4.11 (AA)
Touch target48×48 dp Android, 44×44 pt iOS. WCAG's floor is 24×24 CSS px (SC 2.5.8), but platform guidance is the real bar on phonesMaterial Design; Apple HIG
Text resizeContent still works at 200% text size — no clipping, no overlap, no lost actionsWCAG 2.2 SC 1.4.4
FlashingNothing flashes more than 3 times per secondWCAG 2.2 SC 2.3.1
Colour aloneState (error, selected, required) is never shown by colour only — add an icon, text, or shapeWCAG 2.2 SC 1.4.1
NamesEvery interactive element has an accessible name a screen reader speaksWCAG 2.2 SC 4.1.2

Contrast is computed, not eyeballed: take the two hex values and run the WCAG relative-luminance formula, or a contrast checker. Report the actual ratio ("#8A8A8A on #FFFFFF = 3.5:1, needs 4.5:1").

Screen readers — what to look for (Flutter terms, same idea everywhere):

  • IconButton with no tooltip → announced as just "button". Add tooltip.
  • GestureDetector / InkWell used as a button → no button role. Wrap in Semantics(button: true, label: ...) or use a real button widget.
  • Image with meaning but no semanticLabel; decorative images not excluded (excludeFromSemantics: true).
  • Custom toggles/checkboxes with no toggled / checked semantics.
  • Reading order that jumps around because of Stack / absolute positioning.

Web: the same checks as aria-label, role="button", alt. React Native: accessibilityLabel, accessibilityRole. SwiftUI: .accessibilityLabel.

2. Text scaling and small screens

Test these two conditions, or mark them "not rendered":

  • Text at 2.0×. In Flutter: MediaQuery(data: MediaQuery.of(context).copyWith(textScaler: TextScaler.linear(2.0)), child: ...).
  • 320 dp wide. The smallest common phone width.

What breaks, in order of how often:

  • Fixed-height containers around text → text clipped. Let height grow, or use minHeight.
  • Row with two Texts and no Expanded/Flexible → overflow stripes.
  • Hardcoded widths (width: 400) → off-screen on small phones.
  • Buttons whose label wraps and pushes the icon out.
  • maxLines: 1 on anything a user types (names, titles) with no overflow: TextOverflow.ellipsis.

Clamp text scaling only as a last resort and never below the platform's accessibility sizes — clamping is how apps quietly fail low-vision users.

Responsive layout: branch on the available width, not the device type.

LayoutBuilder(
  builder: (context, constraints) {
    // Material window-size classes: compact < 600, medium < 840, expanded ≥ 840
    if (constraints.maxWidth >= 600) {
      return Row(children: [sidebar, Expanded(child: content)]);
    }
    return content;
  },
)

3. Dark mode

  • Every colour comes from the theme (Theme.of(context).colorScheme, or the project's tokens). A literal Colors.white / Color(0xFF...) in a widget is a finding — it will be wrong in one of the two modes.
  • Check both modes for contrast. A colour that passes 4.5:1 on white often fails on near-black.
  • Avoid pure black #000000 surfaces behind pure white text for long reading — the halo is tiring. Material's dark baseline surface is #121212.
  • Shadows barely show on dark surfaces. Elevation in dark mode is expressed with lighter surface tones, not bigger shadows.
  • Images and illustrations with white backgrounds glow in dark mode. Check them.

4. Motion

  • Respect reduced motion. In Flutter, MediaQuery.disableAnimationsOf(context) is true when the user turned animations off. Non-essential motion (parallax, auto-playing loops, big transitions) must stop or become a fade.
  • Durations by size: small UI feedback (a toggle, a ripple) 100–200 ms; element transitions 200–300 ms; full-screen transitions up to ~500 ms. Anything longer makes the user wait for decoration.
  • Performance: animate with AnimatedBuilder / implicit animations so only the animated subtree rebuilds; put a RepaintBoundary around a continuously animating widget that sits inside a static screen.
  • No flashing over 3 per second (see §1).

5. The four states

Every screen that loads data needs all four. Find the code for each, or prove it is missing.

StatePassFail
LoadingSkeleton or placeholder shaped like the content; the rest of the screen still usable where possibleA lone spinner in the middle of a blank screen for more than a moment
EmptySays why it is empty and gives one next action"No data" / blank space
ErrorPlain-language message + a retry; the error is shown, not only loggedA catch that only logs, leaving the user on a spinner or a blank list
SuccessConfirms the action with the same verb as the button ("Save" → "Saved")Nothing changes, so the user taps again
if (isLoading) return const ContentSkeleton();
if (error != null) return ErrorState(message: error!, onRetry: load);
if (items.isEmpty) return EmptyState(onCreate: openCreate);
return ContentList(items: items);

6. Pressure tests

State each as checked or not checked:

  • A 40-character title and a 30-character username
  • Text at 2.0× and a 320-wide screen
  • Dark mode and light mode
  • No network, and a slow network (does loading ever end?)
  • An empty account (no content, no history)
  • Screen reader on: can you complete the screen's main action by ear alone?

Output Format

# UI/UX Review — <screen> — <date>

Rendered: <which pressure tests you actually ran> | Not rendered: <the rest>

## ♿ Accessibility
### A-1 <title>
**Where:** `path/to/file:LINE`
**Rule:** <standard + number, e.g. WCAG 1.4.3, 4.5:1>
**Measured:** <e.g. 3.1:1, 36×36 dp, no label>
**Who is hurt:** <e.g. low-vision users, screen-reader users>
**Fix:** <exact change>

## 📱 Text scaling & small screens
## 🌙 Dark mode
## 🎬 Motion
## ⏳ States (loading / empty / error / success)
<same shape>

## ✨ Polish (no standard broken)
<taste-level suggestions, clearly separated>

## ✅ Checked and fine
<what you checked that passed — name the check>

Red Flags — stop and fix the review

  • A finding with no number or no source → move it to Polish
  • "Contrast looks low" with no ratio → compute it
  • Passing text scaling without rendering it
  • A colour fix that hardcodes a new hex instead of using the theme
  • Reviewing only the happy path of a screen that has a catch block

Quick Reference

CheckNumber
Normal text contrast4.5:1
Large text / icons / borders3:1
Touch target48×48 dp (Android), 44×44 pt (iOS)
Text scale to test200%
Narrowest width to test320 dp
Flashing limit3 per second
Small feedback motion100–200 ms
Screen transition≤ ~500 ms

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.