mobile-app-design
Skill2 starsChanged 4 months agoMITView on GitHub
Design and review mobile app UI — screens, navigation, bottom sheets, gestures, swipe interactions, empty/loading states, color systems and dark mode. Works for SwiftUI, Jetpack Compose, React Native, Expo and Flutter. Use when building a new screen or flow, when polishing an app that looks generic, cramped or AI-generated, or when asked to "design this screen", "improve the app UI", "make this look professional", "review my mobile design", or "why does this feel cheap".
5 other projects use a version of this.
Adopted by
Most recently taken up by
- fluxnative-app/fluxnative-skillstheir own version, first seen Nov 16, 2026
- podo/design-agent-skillstheir own version, first seen Nov 8, 2026
- fluxnative-app/fluxnative-claudetheir own version, first seen Nov 8, 2026
- jasonfesta/design-agentstheir own version, first seen Oct 15, 2026
- noestradamus/skillstheir own version, first seen Oct 1, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- podo/design-agent-skills★ 32same name
- jasonfesta/design-agents★ 1same name
- noestradamus/skills★ 1same name
- fluxnative-app/fluxnative-claude★ 0same name
- fluxnative-app/fluxnative-skills★ 0same 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 repository2 other files it gives its agents.
- dashboard-designSkill
- landing-page-designSkill
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 9 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
9
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Run these in order. Do not jump to visuals before step 2 is settled — most bad mobile UI is a structure problem wearing a color problem's …
- Distinctive: line 31
Ask what the user is trying to *do* on this screen, and design the shortest path to it.
- Distinctive: line 35
Write the intent down in one sentence before you write any code.
- Distinctive: line 53
Too many top-level destinations?: Don't cram — turn the nav into a full page that *becomes* the home screen (the Notion pattern), which …
- Very specific: line 90
Do not pick colors ad hoc, and never accept a model's first palette — they trend bright, saturated and mutually hostile.
- Distinctive: line 107
Every interactive element needs: default, pressed, disabled (hover does not exist on touch), plus loading where an action takes time.
- Distinctive: line 111
Every screen needs its empty states — and there are two different kinds:
- Very specific: line 137
Run references/audit-checklist.md.
- Distinctive: line 143
Only four things are worth refusing to negotiate.
- Distinctive: line 146
≥ 44pt touch targets: Always. This is the most common failure in generated mobile UI.
- Distinctive: line 151
No emoji as interface icons: Use a real set (Phosphor, Lucide, SF Symbols, Material Symbols) at one consistent stroke weight.
- Very specific: line 156
When reviewing a screenshot — yours or a model's output — grade it against these eight.
“Common” is measured: kept by at least 6 in 10 of the 5 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.
Files
Mobile App Design
Ship mobile UI that looks deliberate instead of generated. This skill is a working method plus a critique vocabulary — use it both to build and to audit.
Polish is what happens when persistence and taste meet the science of design. Agents have no eyes. You can only compensate with a checklist and a screenshot.
When to use which mode
| Situation | Mode |
| New screen or flow from scratch | Build — run the pipeline below in order |
| Existing UI feels off, generic, cramped | Audit — go straight to references/audit-checklist.md |
| One narrow question (color? gesture?) | Read the matching reference file only |
Build pipeline
Run these in order. Do not jump to visuals before step 2 is settled — most bad mobile UI is a structure problem wearing a color problem's clothes.
1. Intent before interface
Ask what the user is trying to do on this screen, and design the shortest path to it. A rental app's intent is "find a place" → the search field is the design. The hero image and the wordmark are decoration you add afterwards, if at all.
Write the intent down in one sentence before you write any code. If you cannot, the screen is doing too many things.
One screen does one thing. Settings is settings. The editor is the editor. The home screen is the only routine exception. When you need to add something new, reach for a different screen, not a busier layout.
2. Structure and navigation
Pick the navigation shell first — it constrains everything else. See
references/layout-navigation.md for the full decision tree.
Hard constraints:
- Bottom bar: 5 links is the absolute limit, 3–4 is right. More than that means your information architecture is wrong, not that you need smaller icons.
- Every touch target ≥ 44pt. People have fat fingers. This is not negotiable and it is the single most common failure in agent-generated mobile UI.
- Too many top-level destinations? Don't cram — turn the nav into a full page that becomes the home screen (the Notion pattern), which frees the bottom of the screen for a large search field or primary action.
3. Layout: one direction per section
Desktop lays out content in two directions at once — columns and rows. Mobile gets one direction per section. Either stack items vertically, or scroll them horizontally off the edge. Never both in the same section.
This one rule converts most desktop/dashboard layouts to mobile correctly.
Four building blocks exist, ignoring floating menus: cards, text/links, images, inputs. Everything is a composition of those.
- Never double-nest cards. Padding on padding cramps everything. Group with whitespace instead of a second container — whitespace is a container you don't have to pay padding for.
4. Type and space
Counterintuitive but load-bearing: type gets larger on mobile, not smaller. iOS base font size is 17px; macOS is 13px. Squishing desktop type down to fit more on a small screen is the beginner tell.
- One sans-serif family. You will essentially never need a second one.
- Large headings: tighten letter-spacing to -2% to -3% and line-height to 110–120%. This single adjustment makes big text read as professional.
- Space on a 4pt grid — not because 4 is magic, but because everything stays halvable, which is what produces consistency.
- Group related items with tighter spacing; separate groups with more. Proximity is hierarchy you get for free.
Full detail in references/foundations.md.
5. Color
Do not pick colors ad hoc, and never accept a model's first palette — they trend
bright, saturated and mutually hostile. Build the four layers in order: neutral
foundation → functional accent → semantic → theming. See references/color.md.
The three rules that catch the most damage:
- Backgrounds stay in the background. Neutral, never a bright gradient.
- Icons need no color. Their job is to be a recognizable symbol. Color is reserved for communicating status.
- Dark mode is not inverted light mode. Dark values need roughly double the distance between steps to read as different, and surfaces get lighter as they elevate — there are no shadows to do that work for you.
6. States — the actual product
Designing only the happy path is what separates a mockup from a product.
Every interactive element needs: default, pressed, disabled (hover does not exist on touch), plus loading where an action takes time. Inputs additionally need focus and error with a message.
Every screen needs its empty states — and there are two different kinds:
- First run: full-screen, points at the primary action, one popover explaining how it works. Do not fill it with fake cards inviting the user to do six things.
- No results: imagery, acknowledge the query explicitly, offer suggestions in case of a typo, and give an action to exit the state.
See references/states-feedback.md.
7. Motion and gestures
Motion is where mobile stops feeling like a website in a phone-shaped box.
- Never linear easing. Real things accelerate and decelerate. The curve carries the tone: snappy, springy, or slow and smooth.
- Motion follows the direction of the swipe. Swipe up → content rises in. Swipe back → the background moves left ~35% and animates right.
- Always pair a gesture with a visible control. Gmail lets you swipe to delete and tap into a menu. Gestures are for the initiated; buttons are for everyone.
- Slide-to-confirm for irreversible actions. Purchases, sends, deletions. It is much harder to slide a slider by accident than to tap a button.
See references/motion-gestures.md.
8. Audit before you call it done
Run references/audit-checklist.md. If you can render or screenshot the screen,
look at it and critique it against the eight principles — you cannot evaluate
balance and whitespace from source code alone.
Hard rules
Only four things are worth refusing to negotiate. Everything else in this skill is guidance — apply judgement.
- ≥ 44pt touch targets. Always. This is the most common failure in generated mobile UI.
- Destructive is red. Even when red is nowhere in the brand palette. A delete button in your brand color is a usability failure, not a brand win.
- Never ship a screen without its empty and loading states.
- No emoji as interface icons. Use a real set (Phosphor, Lucide, SF Symbols, Material Symbols) at one consistent stroke weight.
The critique vocabulary
When reviewing a screenshot — yours or a model's output — grade it against these
eight. They are the ingredients of "I know it when I see it"; detailed treatment in
references/foundations.md, applied rubric in references/audit-checklist.md.
| Principle | Failing looks like |
| Contrast | No obvious focal point; everything competes equally |
| Hierarchy | Wordmark, tagline and card titles fight for the same attention |
| Alignment | Elements don't share edges or axes |
| Proximity | Related things not grouped; unrelated things jammed together |
| Repetition | Three type weights, two icon styles, four corner radii |
| Balance | Bottom-heavy or lopsided; one dense block dominates |
| Whitespace | Every pixel working; nothing breathes |
| Unity | Reads as a template with content poured in |
Unity is what emerges when the other seven agree. If a screen reads as generic, one of them is failing — find which.
Treat generated UI as a base to build on, never as the deliverable. The
30-second tell list lives at the top of references/audit-checklist.md.
Reference files
references/foundations.md— the eight principles, hierarchy, spacing, type, icons, shadows, signifiersreferences/color.md— the four-layer color system, dark mode, semantic color, OKLCH for charts and themingreferences/layout-navigation.md— navigation shells, bottom sheets, contextual actions, cards, screen compositionreferences/motion-gestures.md— easing, the three classes of swipe interaction, transitions, long-press, micro-interactionsreferences/states-feedback.md— element states, empty states, loading, errors, toasts, onboarding, optimistic UIreferences/reference-ui.html— a working screen implementing the system (color ramps, type scale, states, components). Open it when you need concrete structure rather than descriptionreferences/audit-checklist.md— the full review rubric, including the 30-second tell list
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
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.

