RomM v2: Universal Input & Universal Viewport

SkillDev tools

Guides your agent in adding keyboard, mouse, touch, and gamepad input plus responsive layout to the RomM v2 frontend.

Use RomM v2: Universal Input & Universal Viewport in Claude, ChatGPT or Ahel Desktop

Free. Sign in, add RomM v2: Universal Input & Universal Viewport and connect your AI. About a minute.

Also: Claude Code · Cursor · Codex

Then ask your AI: use the RomM v2: Universal Input & Universal Viewport skill

Details

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

RomM v2: Universal Input & Universal ViewportStart free
About this skill

Universal input (mouse, touch, keyboard, gamepad) and responsive/universal-viewport layout in the RomM v2 frontend. Use when adding interactive v2 components, focus management, spatial navigation, gamepad/keyboard handling, modality-gated focus rings, breakpoints, or responsive layout. Covers useInp

What this skill tells your AI

The instructions your AI receives, as published by rommapp/romm in .claude/skills/frontend-v2-input/SKILL.md and read by ahel’s review.

Premise: all v2 UI works with mouse, touch, keyboard, and gamepad; and every surface reads cleanly from a 320px phone to a 4K display. Both mechanisms are fixed; don't invent a parallel one.

The input system lives in src/v2/composables/useInput/ (bus, keyboard, gamepad, actions, scope). It generalises the original gamepad-only src/console/ system. There are no /console/* routes in v2.


Modality

useInputModality sets data-input="mouse|touch|key|pad" on <html> from the most recent input.

  • Focus rings appear only with key and pad (CSS in global.css). Never use bare :focus in styles; use the modality-gated selectors, or focus rings flash on mouse click.
  • Modality gates appearance and behaviour, never size: focus rings, hover-reveal, autofocus. A tap is followed by compatibility mouse events, so anything sized off data-input resizes under the finger. Hit targets scale by breakpoint instead (below).

Coverage: every interactive primitive participates

Buttons, list items, tabs, menu items, focusable cards, toggleable chips: all participate in spatial navigation (not optional). A new interactive primitive must:

  • be focusable (a natively focusable element, or a proper tabindex);
  • react to logical actions (confirm/cancel) from useInput, in addition to native click;
  • show a modality-gated focus state.

Storybook play() covering gamepad input is required only when applicable (the primitive is interactive enough that gamepad navigation matters).

Focus geometry

Each view declares its layout with focus primitives: RFocusZone, RFocusGrid, RFocusRow, RFocusColumn. Multiple regions = multiple zones. Predictable up/down/left/right movement is the view's responsibility.

Element-level global shortcuts

Above per-view geometry, some elements bind globally regardless of focus location:

  • UserMenu opens on Start.
  • Navbar tabs cycle with LB/RB (plus D-pad).
  • Context menu opens on X or Y.

Scope (overlay stack)

When a dialog opens, push a scope; when it closes, pop. This stops Escape from closing two things at once and stops confirm leaking to controls beneath an overlay. RDialog and RMenu manage their scope automatically; custom overlays are an anti-pattern; go through the primitives.


Responsive layout (universal viewport)

  • Single breakpoint source: useBreakpoint (src/v2/composables/useBreakpoint/). Material thresholds: xs <600, sm 600–959, md 960–1279, lg 1280–1919, xl ≥1920. installBreakpointAttribute() mirrors the active set onto <html data-bp="…"> (from AppLayout and AuthLayout in the app, and from .storybook/preview.ts so stories match production CSS).
  • Layout switches live in CSS via the attribute selector: html[data-bp~="xs"] .foo { … }, html[data-bp~="sm-and-down"] .foo { … }. No raw @media for layout: the only allowed @media are prefers-reduced-motion and print. The attribute is on <html> so it reaches teleported overlays. romm/no-layout-media-query (ESLint) enforces this in SFC styles.
  • Conditional rendering (mount/unmount a different component per tier) uses the useBreakpoint() refs in <script>, e.g. v-if="xs". Prefer mount-gating over display:none for focusable chrome so hidden controls never sit in the tab/spatial-nav order.
  • --r-row-pad is the global horizontal gutter, already re-scoped responsive in global.css (36 → 20 → 14px). Consume var(--r-row-pad); don't hard-code a smaller xs padding per component.
  • Touch targets scale with the breakpoint, never with data-input: --r-touch-target (44px) at sm-and-down, less only where the row would not fit.
  • Overlays go full-bleed on xs. RDialog renders full-screen / bottom-sheet on phones (fullscreenOnMobile, default on); RMenu bottom-sheets large menus. Never float a 600px dialog on a 360px screen.
  • Label→icon collapse (the AppNav precedent) is the canonical way to compress chrome; the four primary destinations relocate to BottomNav on sm-and-down.
  • Grids size via useResponsiveColumns (ResizeObserver), never a fixed column count.

Verification adds a breakpoint sweep; see review-polish. In Storybook, use the viewport toolbar presets in .storybook/rommViewports.ts (RomM xs/sm/md tiers, Steam Deck for lg, AYN Thor top/bottom screens). A story that needs a tier sets globals: { viewport: { value: "rommPhoneXs" } } instead of writing data-bp itself. Layout that depends on html[data-bp~="xs"] needs a canvas under 600px wide (e.g. 390×844 · RomM phone), not just useBreakpoint() switching subtabs at tablet widths.

Signals

GitHub stars
13k
Forks
747
Last commit
Oct 2026
Advanced
Item type
skill
Key
frontend-v2-input
Source
github.com/rommapp/romm