Moai Ref Ui Polish logo

Moai Ref Ui Polish

OrganizationPopular
modu-ai
moai-ref-ui-polish

UI polish and interface-completion reference: the small visual details — concentric border radius, optical alignment, shadow-vs-border, motion easing, typography smoothing, tabular numbers, icon stroke weight, hit areas — that separate polished interfaces from generic ones. Agent-extending skill that amplifies frontend/UI domain work with production-grade "interface taste" rules. NOT for: backend logic, database design, DevOps, security audits, non-UI work.

Overview

Publishermodu-ai
Repositorymoai-adk
Skill namemoai-ref-ui-polish
Stars
1.2K
Forks
226
Bundled files
2
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 2 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by modu-ai on GitHub. Read the source before you install it.

Installation

Install the Moai Ref Ui Polish AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/modu-ai/moai-adk/tree/main/.claude/skills/moai-ref-ui-polish
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/modu-ai/moai-adk.git /tmp/moai-adk
mkdir -p .claude/skills
cp -r /tmp/moai-adk/.claude/skills/moai-ref-ui-polish .claude/skills/moai-ref-ui-polish
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Moai Ref Ui Polish in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Moai Ref Ui Polish on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Moai Ref Ui Polish is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

UI Polish Reference

Target Agents

  • manager-develop - Applies polish rules during frontend/UI component implementation (cycle_type=tdd or cycle_type=ddd context)
  • /moai review - UI-design review surface; equivalently available as a per-spawn Agent(general-purpose) frontend specialist per archived-agent-rejection.md §C

Core Philosophy

Great interfaces are a collection of small details that compound into a great experience. AI agents frequently miss these details — an ease-in easing on an enter animation (should be ease-out), a solid border where a semi-transparent shadow reads better, or mismatched radii on nested elements. None of these is catastrophic; together they separate "polished" from "generic".

Before suggesting polish changes, identify the project's existing styling system (design tokens, spacing scale, motion library). Never introduce a second styling system for polish fixes — extend the existing one.

Geometry and Alignment

PrincipleRuleRationale
Concentric Border RadiusouterRadius = innerRadius + paddingMismatched radii on nested elements is the most common cause of an interface "feeling off"
Optical over Geometric AlignmentWhen geometric centering looks wrong, align opticallyButtons with icons, play triangles, and asymmetric icons need a manual nudge; geometric center is visually off-center for these shapes

Elevation and Structure

ElementUseAvoid
Semi-transparent layered box-shadowDepth, elevation, floating surfacesSolid borders for depth (they read heavy and flat)
BordersStructure, dividers, separators, selected/focus stateUsing shadows for structural separation (ambiguous)

Motion

PatternRuleCommon Mistake
Enter animation easingease-out (decelerate) — element arrives calmlyease-in on enter (element appears to slam into place)
Exit animation easingease-in (accelerates away), mirroring the enter curve (small fixed translateY)Full-height collapse, or harsher-than-enter motion
Interruptible state changesCSS transition (can be interrupted mid-animation)keyframes for interactive states (cannot interrupt)
Staged entrancesStagger semantic chunks ~100ms — only for infrequent staged entrancesStaggering routine, high-frequency interactions (feels sluggish)
Contextual icon animationopacity/scale/blur cross-fade (scale 0.25→1, opacity 0→1, blur 4px→0)Toggling visibility (jarring, no transition)
Press feedbackscale(0.96) on click — always 0.96Smaller than 0.95 (reads as a bug, not a press)
First-render enterSkip with initial={false} on AnimatePresence (or equivalent)Enter animation fires on page load (disorienting)
Motion restraintNo custom animation on high-frequency interactions; motion is never the only feedback channelAnimating everything (noise, performance cost, accessibility)

Motion Easing Values (Web/CSS)

CaseValue
With motion library (Framer Motion et al.)transition: { type: "spring", duration: 0.3, bounce: 0 }
Without motion library (CSS)cubic-bezier(0.2, 0, 0, 1) for the standard "decelerate" curve
Nevertransition: all — always specify exact properties (transition-property: scale, opacity)

Motion Accessibility and Cost

RuleDetail
Reduced-motion branch (required)Every non-decorative animation needs a prefers-reduced-motion: reduce path. Vestibular-disorder users are physically harmed by large-displacement and parallax motion. Reduce to an opacity cross-fade or remove the animation — never merely shorten it. The branch is authored at the same time as the animation, not retrofitted
Animate the compositor, not the layouttransform and opacity are composited and skip layout and most paint. width, height, top, left, margin, and padding trigger layout on every frame (thrashing), which is why the same visual effect janks when driven by the wrong property. Move with translate, resize with scale

Motion principles behind these rules — the three decision passes, motion layers, the 1/3 rules, stagger budgets, personality archetypes, Disney-adapted ranges, and the emotion-to-motion map — live in references/motion-principles.md (L3, load on demand when motion is the substance of the work).

Typography

RuleImplementation (Web/CSS)When
Font smoothing-webkit-font-smoothing: antialiased on root layoutmacOS targets (sharper rendering)
Tabular numbersfont-variant-numeric: tabular-numsDynamically updating numbers (counters, timers, prices) — prevents layout shift
Heading wrappingtext-wrap: balanceHeadings (prevents orphan words, evens line lengths)
Body wrappingtext-wrap: prettyBody paragraphs (avoids orphan widows)

Imagery

RuleValue
Image outline1px subtle outline at low opacity
Outline color (light mode)pure black — oklch(0 0 0 / 0.1)
Outline color (dark mode)pure white — oklch(1 0 0 / 0.1)
NeverTinted neutral outlines (read as a color choice, not a separation cue)

Interaction

RuleValue
Minimum hit area (touch/mobile)44 × 44 px
Minimum hit area (dense desktop)40 × 40 px (extend with pseudo-element if the visual is smaller)
Hit area overlapNever let hit areas overlap
Hover on touch (never load-bearing):hover does not exist on touch — anything reachable only by hover is simply gone on a phone. Keep affordances visible by default and gate hover styling behind @media (hover: hover) and (pointer: fine)
Hover on pointer (always present)The inverse rule: on a pointer device every clickable surface needs a distinct hover state, transitioning over ~100-200ms. A pointer resting on a target with no feedback reads as broken
will-changeOnly transform, opacity, filter — and only when first-frame stutter is observed; never will-change: all

Icons

RuleDetail
Stroke matches text weight1.5px stroke beside regular (400) text; 2px beside semibold (600) text
One stroke weight per setNever mix icon libraries in one interface
State via color, not assetsIcons use currentColor; states come from CSS color/opacity — never separate asset files
Outline vs fillOutline variant is default; fill variant marks the active state

Review Modes

ModeCoverageFinding Cap
quickPrimary user path, high-traffic states; HIGH/MEDIUM issues only5
fullEntire scope across all categories (Typography, Surfaces, Motion, Icons, Interaction)15

Severity

LevelMeaning
HIGHMakes interaction inaccessible, misleading, unreadable, or repeatedly disruptive
MEDIUMNoticeable usability or consistency problem
LOWIsolated polish issue (full mode only)

Verdict

VerdictCondition
BlockAny HIGH finding remains
Needs changesOnly MEDIUM or LOW findings remain
ApproveNo actionable findings remain

Auditing an existing codebase rather than reviewing a diff? references/design-audit.md (L3, load on demand) carries the mechanical detection patterns behind the checklists below — motion gaps, accessibility violations, layout-property animation, and the duration/easing inventories that only surface across a whole codebase. It reports into the severity scale and finding caps above, not a separate one.

Common Rationalizations

RationalizationReality
"The framework handles animations correctly by default"Framework defaults are generic. Easing, timing, and stagger must be specified per interaction, not inherited.
"A solid border is simpler than a layered shadow"A solid border communicates structure; depth needs transparency. Using a border for elevation reads heavy and flat.
"Geometric centering is correct, optical adjustment is bikeshedding"Geometric center is visually off for asymmetric shapes (icons, play triangles). Optical alignment is the difference users feel but cannot name.
"Motion is optional polish, the interface works without it"Motion is feedback. Without it, state changes are ambiguous (did it register?). But motion must be restrained — animating everything is noise.
"Hit area equals visible area"A 20px icon needs a 44px hit area on touch. Overlapping or undersized hit areas are the most common mobile usability defect.
"transition: all is convenient"It transitions properties you did not intend (layout, color), causing unexpected animation. Always specify exact properties.

Red Flags

  • Enter animation uses ease-in (should be ease-out)
  • Nested elements with mismatched border radii (no concentric relationship)
  • Solid border used for depth/elevation instead of semi-transparent shadow
  • transition: all instead of specific properties
  • Hit area smaller than 40×40px, or overlapping hit areas
  • Dynamically updating numbers without tabular-nums (layout shifts on each update)
  • Icon stroke weight mismatched with adjacent text weight
  • Separate icon asset files per state instead of currentColor recoloring
  • Page-load enter animation fires without a skip (initial={false})
  • will-change: all, or will-change set permanently instead of only on stutter
  • Animation with no prefers-reduced-motion branch (or a branch that only shortens the duration instead of removing the displacement)
  • Movement driven by top/left/width/height where a transform would do (layout thrashing every frame)

Verification

  • Enter animations decelerate (ease-out); exit animations accelerate (ease-in), and entrances run longer than exits
  • Every non-decorative animation has a prefers-reduced-motion: reduce branch
  • Movement uses transform/opacity, not width/height/top/left
  • Nested elements follow outerRadius = innerRadius + padding
  • Depth uses semi-transparent shadow; structure uses border
  • Every transition specifies exact properties (no transition: all)
  • Hit areas meet 44×44px (touch) / 40×40px (desktop) minimum, no overlap
  • Dynamic numbers use tabular-nums; headings use text-wrap: balance
  • Icons use one stroke weight matching text; states via currentColor, not separate assets
  • First render skips enter animation (initial={false} or equivalent)
  • High-frequency interactions have no custom animation; motion is not the only feedback

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Moai Ref Ui Polish AI skill do?

UI polish and interface-completion reference: the small visual details — concentric border radius, optical alignment, shadow-vs-border, motion easing, typography smoothing, tabular numbers, icon stroke weight, hit areas — that separate polished interfaces from generic ones. Agent-extending skill that amplifies frontend/UI domain work with production-grade "interface taste" rules. NOT for: backend logic, database design, DevOps, security audits, non-UI work.

Why use Moai Ref Ui Polish on TypingMind?

Because you install it once and use it with any model. Moai Ref Ui Polish is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Moai Ref Ui Polish in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/modu-ai/moai-adk/tree/main/.claude/skills/moai-ref-ui-polish. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Moai Ref Ui Polish?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Moai Ref Ui Polish?

As many as you like. As long as a model supports skills, you can use Moai Ref Ui Polish with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Moai Ref Ui Polish AI skill free?

Yes. It is published on GitHub by modu-ai under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇