better-ui
jakubkrehel/skills · Agent Skill
Design engineering principles for making interfaces feel polished. Use when building UI components, reviewing frontend code, implementing animations, hover states, shadows, borders, micro-interactions, enter/exit animations, or any visual detail work. Triggers on UI polish, design details, "make it feel better", "feels off", stagger animations, border radius, optical alignment, image outlines, box shadows.
Jump to install ↓Overview
| Publisher | |
|---|---|
| Repository | jakubkrehel/skills ↗ |
| Skill folder | better-ui |
| Stars | 8,643 |
| Forks | 330 |
| Bundled files | 8 |
| License | MIT |
| Last updated | October 25, 2026 |
| Links | Skill source ↗ · Repository ↗ |
Mozgi check
What we found in the skill files. Details come with a subscription.
Installation
A skill is a folder with a SKILL.md file. Pick your agent and run the commands in a terminal (macOS, Linux or Git Bash on Windows) from your project root.
Copy the prompt into Claude, Codex or another assistant: it will review the skill and install it for you.
Install the “better-ui” skill from mozgi.io (https://mozgi.io/skills/jakubkrehel-better-ui/) for me. 1. Download the archive: https://mozgi.io/skills/jakubkrehel-better-ui/download If it fails, take the skill folder from the source: https://github.com/jakubkrehel/skills/tree/main/skills/better-ui 2. Read SKILL.md and the other files. Check the skill is safe: no hidden commands, no sending data elsewhere, no scripts doing something other than described. If anything looks suspicious, stop and tell me. 3. Put the “better-ui” folder into your agent's skills folder: .claude/skills/ for Claude Code, .agents/skills/ for Codex, .cursor/skills/ for Cursor, .gemini/skills/ for Gemini CLI. Inside the project if we are in one, otherwise in the home folder (~/.claude/skills/ etc.). 4. Make sure <skills folder>/better-ui/SKILL.md exists. 5. Briefly tell me what this skill does and offer to try it on my task. If you cannot do a step yourself, tell me what to click or type.
git clone --depth 1 --branch main https://github.com/jakubkrehel/skills.git /tmp/skills mkdir -p .claude/skills cp -r /tmp/skills/skills/better-ui .claude/skills/better-ui
Restart Claude Code — the skill appears in the list; call it manually with /better-ui. For all projects use ~/.claude/skills.
git clone --depth 1 --branch main https://github.com/jakubkrehel/skills.git /tmp/skills mkdir -p .agents/skills cp -r /tmp/skills/skills/better-ui .agents/skills/better-ui
Codex looks for skills in .agents/skills from the current folder up to the repo root; call one with /skills or $better-ui. For all projects use ~/.agents/skills.
git clone --depth 1 --branch main https://github.com/jakubkrehel/skills.git /tmp/skills mkdir -p .cursor/skills cp -r /tmp/skills/skills/better-ui .cursor/skills/better-ui
Cursor picks the skill up on start; call it with “/” in Agent chat. Cursor also reads .claude/skills and .agents/skills. For all projects use ~/.cursor/skills.
git clone --depth 1 --branch main https://github.com/jakubkrehel/skills.git /tmp/skills mkdir -p .gemini/skills cp -r /tmp/skills/skills/better-ui .gemini/skills/better-ui
Gemini CLI offers to activate the skill when a task matches — you approve it. For all projects use ~/.gemini/skills.
Read SKILL.md and any companion files before deciding whether to install. Click a file to preview it.
SKILL.md contents
Original author text · read-onlyUI polish
This skill holds the visual polish for surfaces, icons and motion, with the exact value each detail takes. It applies once the underlying interaction is sound, and a polish finding never outranks a broken interaction.
Exact values, optional polish
The values below are exact, not ranges to approximate. cubic-bezier(0.2, 0, 0, 1) is not cubic-bezier(0.4, 0, 0.2, 1), and 0.96 is not 0.95. The optical nudges and the concentric padding cutoff are the exceptions. They are starting points, judged by eye.
Keep the project's component library, tokens and density, and match its motion language wherever no rule here gives a value. A deliberate and consistent project convention, such as a style with no shadows, is a preference and not a finding. The same detail done two ways within the project is a finding.
Text wrapping, font rendering, tabular numbers and text spacing belong to better-typography. Hit areas, keyboard support, ARIA and the reduced-motion requirement belong to better-accessibility. Grouping, section spacing, breakpoints and spatial RTL belong to better-layout, except directional icon mirroring. Color tokens and contrast measurement belong to better-colors.
Outer radius equals inner radius plus padding
Where nested surfaces share a visible, even inset, the outer radius is the inner radius plus the padding plus any border width. Past 24px of padding, or where the padding is deliberately asymmetric, treat the layers as separate surfaces and keep each one's radius token. Recipes are in surfaces.md ↗.
Align optically where geometry looks off
Where geometric centering looks off, nudge by eye. Give a button 2px less padding on its icon side, shift a play triangle toward its point and fix asymmetric glyphs in the SVG itself. Recipes are in surfaces.md ↗.
Shadows for elevation, borders for structure
Where a border exists only to create depth, replace it with layered transparent box-shadow values. Keep borders on dividers, separators, table cells and selected states. Keep them on form inputs too, whose boundary needs 3:1 non-text contrast under better-accessibility. Focus rings belong to better-accessibility as well.
Forced-colors mode removes every box-shadow. Keep border: 1px solid transparent under a shadow ring so that mode still draws an edge. Recipes are in surfaces.md ↗.
Outline images in pure black or white
Give content images a 1px outline inset by 1px. Use oklch(0 0 0 / 0.1) in light mode and oklch(1 0 0 / 0.1) in dark. Never use a palette near-black, a tinted neutral or the accent color, because the tint shows as a colored fringe on the image edge. Skip transparent artwork such as logos and illustrations. The recipe is in surfaces.md ↗.
Transitions, not keyframes, for interactive state
Drive interactive state changes with CSS transitions or a motion library's springs, which retarget when the user reverses mid-flight. Keyframes run a fixed timeline and cannot reverse, so reserve them for sequences that run once. See animations.md ↗.
Press scales to 0.96
A pressed button scales to 0.96 over 150ms with ease-out, and a disabled one never scales. Give the button an opt-out where the motion would distract, through the component's existing variant API or a static prop. See recipes for CSS, Tailwind and Motion ↗.
High-frequency interactions respond instantly
Keystrokes, row hovers and tab switches get instant feedback, or a transition of 150ms or less on opacity or background-color. Reserve expressive motion for infrequent moments such as a view's first load, a success state or an empty state.
Every animated state change also leaves a static cue in the form of a color, icon or label. Motion is never the only feedback channel.
Gate motion behind the reduced-motion preference
better-accessibility owns the requirement, and no recipe here ships without it. Run movement, scale and blur only under prefers-reduced-motion: no-preference. Under reduced motion, replace them with an opacity cross-fade rather than removing the element instantly. Recipes for CSS, Tailwind and Motion are in animations.md ↗.
Stagger infrequent entrances by 100ms
Where sequence communicates hierarchy in a staged entrance, split the content into semantic chunks such as title, description and actions. Stagger them 100ms apart, each entering with opacity, 4px of blur and 12px of translateY over 300ms with ease-out. Never stagger routine interactions. See enter-exit.md ↗.
Exits are shorter and smaller than enters
Exit with opacity, 4px of blur and a fixed -12px translateY over 150ms with ease-out, never the full container height. Slide fully out only where the destination carries meaning, such as a drawer closing. Remove the element instantly where motion adds no information. See enter-exit.md ↗.
Skip state animations on first render
Set initial={false} on an AnimatePresence that wraps a state swap, so the default state does not animate in on mount. Never set it around an intentional entrance such as a staggered hero. See animations.md ↗.
Cross-fade contextual icons with exact values
Where an icon swaps on an infrequent state change, such as play to pause or copy to copied, cross-fade it. Scale runs 0.25 to 1, opacity 0 to 1 and blur 4px to 0px. With a motion library, use { type: "spring", duration: 0.3, bounce: 0 }, and bounce is always 0. Without one, keep both icons in the DOM and cross-fade over 300ms with cubic-bezier(0.2, 0, 0, 1).
A tab's icon swap and the actions a row reveals on hover fall under High-frequency interactions respond instantly. Both recipes are in icon-transitions.md ↗.
Suppress transitions on theme switch
Disable every transition for the theme swap, force a style flush and restore transitions after the next frame. Otherwise every color transition fires at once and the switch smears. See the recipe ↗.
Transition only what changes
Name the exact properties, as in transition-property: scale, opacity, and never all. See performance.md ↗.
Name the animated property in will-change
Add will-change only after you see first-frame stutter, and name the property you animate: scale for scale, transform for transform. Limit it to transform properties, opacity and filter, and never use all. See performance.md ↗.
Hover effects only on hover-capable pointers
Put hover-only styling behind @media (hover: hover). On touch, :hover latches after a tap and reads as a stuck selected state. Tailwind 4's hover: variant already compiles under that query, and Tailwind 3's does not. Where a control draws its own pressed state, set -webkit-tap-highlight-color: transparent so the default gray flash does not double it.
Contain scroll inside overlays
Give scrollable dialogs, drawers, menus and side panels overscroll-behavior: contain, so scrolling past their end never scrolls the page behind.
Match icon stroke to text weight
An icon's rendered stroke tracks the weight of the text beside it, from 1.5px at 400 to 2.5px at 700. Use one icon library per surface. The table, sizing and grid rules are in icons.md ↗.
One SVG, recolored per state
Icons use currentColor and take hover, selected and disabled states from CSS color and opacity, never from separate assets. Outline is the default variant, and fill marks the active state. Under dir="rtl", mirror only icons whose meaning follows reading direction. See icons.md ↗.
Before you finish
| Detection | Fix |
| --- | --- |
| Padded parent and child with the same rounded-* or border-radius | Add the padding to the outer radius |
| box-shadow: 0 0 0 1px ring with no border beside it | border: 1px solid transparent |
| outline-slate-*, outline-zinc-* or a hex outline color on img | outline-black/10 and dark:outline-white/10 |
| transition-all or transition: all | Name the changing properties |
| animation: set on :hover, .open or another toggled class | A transition on the same properties |
| active:scale-95, scale-90 or whileTap below 0.96 | 0.96 |
| :active scale with no :disabled exclusion | enabled:active: or :not(:disabled):active |
| bounce: set to anything but 0 | bounce: 0 |
| opacity: 0 in a base rule that only an animation reveals | Hidden state in the from keyframe, animation-fill-mode: both |
| translate, scale, filter or motion.* animation with no prefers-reduced-motion, motion-safe: or MotionConfig guard | Gate it and cross-fade opacity under reduced motion |
| Color transitions plus a theme toggle with no disableTransitionOnChange or override | The theme-switch suppression recipe |
| will-change: all, or will-change on an element that never animates | Name the animated property or delete it |
| fill="#..." or stroke="#..." inside an icon SVG | currentColor |
| Plain-CSS :hover or Tailwind 3 hover: outside @media (hover: hover) | Wrap it in the query |
| overflow: auto or overflow-y-auto on a dialog, drawer or menu with no overscroll-behavior | overscroll-behavior: contain |
Reporting
Severity. HIGH breaks an interaction, as a keyframe toggle that cannot reverse or a hover state stuck on touch does. Two of better-interface's escalation triggers land here and are HIGH on sight. One is motion that ignores prefers-reduced-motion, and the other is a state change carried by motion alone. MEDIUM is a visible inconsistency in surfaces, icons or motion. LOW is isolated polish.
Verification. Without a browser, read every state the component defines from the code, such as hover, pressed, selected, loading and empty, with its durations and easings. With one, walk each state and replay motion at 10% speed in the browser's Animations panel. Report every check you could not run as Not verified.
Format. Group findings under the principle each violates, ordered by severity, one row per root cause listing every location it appears in:
| Severity | Location | Before | After | Why |
| --- | --- | --- | --- | --- |
Location is path/to/file:line. Why names the principle and the user impact.
End with Block when any HIGH remains, Approve otherwise, leaving the rest in the table as work to do. Never Approve coverage you did not inspect. With nothing to report, state "No actionable UI-polish findings" and report verification.
Frequently asked questions
What does the better-ui skill do?
Design engineering principles for making interfaces feel polished. Use when building UI components, reviewing frontend code, implementing animations, hover states, shadows, borders, micro-interactions, enter/exit animations, or any visual detail work. Triggers on UI polish, design details, "make it feel better", "feels off", stagger animations, border radius, optical alignment, image outlines, box shadows.
How do I install better-ui?
Open the “Installation” block on this page and pick your agent — Claude Code, Codex, Cursor or Gemini CLI — for ready-made commands that copy the skill folder to the right place. Or copy the prompt from the “Via assistant” tab and let the assistant review and install the skill.
Which AI models work with better-ui?
Skills are plain Markdown instructions (the Agent Skills format), so better-ui works with any agent that reads SKILL.md: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and more. The model loads the instructions only when a task matches the skill's description.
Is better-ui free?
The skill is published openly by jakubkrehel under the MIT license. The skill itself costs nothing — you only pay your AI provider for the tokens you use.
Is it safe to install skills?
A skill is a set of instructions your agent will follow, so read SKILL.md and its companion files before installing. Scripts in the skill folder only run if your agent runs them — keep an eye on what it does and do not grant unnecessary permissions.
How is a skill different from an MCP server or a plugin?
An MCP server or plugin gives the model new tools — code that runs and returns a result. A skill gives it knowledge and process: how to approach a task, which steps to follow and what good output looks like. It needs no server or API key.