Accessibility As Code logo

Accessibility As Code

Community
zakariaf
accessibility-as-code

Enforces accessibility as a correctness property authored into each widget — Semantics(button/label) or ExcludeSemantics on every node, a11y state read from MediaQuery not app state, never MediaQuery.withClampedTextScaling / textScaleFactor / FittedBox / TextOverflow.ellipsis to fit a label, non-color redundant channels (icon+label+shape+text) for every state, contrast against composited backgrounds (4.5:1 body / 3:1 large), 44px single-tap targets, OrdinalSortKey traversal, and honoring boldText / reduce-motion. Use when adding a GestureDetector/InkWell or any tap target, adding an Icon or Image, reaching for withClampedTextScaling/FittedBox/ellipsis/textScaleFactor to make text fit, encoding state via color, sizing type, ordering focus traversal, or reviewing any View for screen-reader/switch/low-vision support.

Overview

Publisherzakariaf
RepositoryFlutter-Skills
Skill nameaccessibility-as-code
Stars
64
Forks
15
Bundled files
Instructions only
LicenseMIT
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.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by zakariaf on GitHub. Read the source before you install it.

Installation

Install the Accessibility As Code 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/zakariaf/Flutter-Skills/tree/main/skills/accessibility-as-code
  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/zakariaf/Flutter-Skills.git /tmp/Flutter-Skills
mkdir -p .claude/skills
cp -r /tmp/Flutter-Skills/skills/accessibility-as-code .claude/skills/accessibility-as-code
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Accessibility As Code 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 Accessibility As Code 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 Accessibility As Code 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.

Accessibility is a coding standard, not a checklist

Accessibility is authored into the widget as you write it — the same tier as "the button actually works." There is no lint for any of this: flutter_lints and very_good_analysis ship zero a11y rules. Enforcement is the widget itself, review of your own diff, and widget tests. Applies to every interactive View, every Icon/Image, every state, and every color choice.

Testing mechanics (guideline matchers, golden lanes, RTL, honest conformance limits) live in widget-golden-and-a11y-testing; this skill is the authoring discipline.

Non-negotiable rules

  1. Every interactive node gets Semantics(button: true, label: ...) — a raw GestureDetector with no semantics silently locks out every screen-reader and switch user. Correct semantics make TalkBack, VoiceOver, Switch Access and Switch Control work for free; it is definition-of-done, never a backlog ticket.
  2. Every Icon/Image gets a semanticLabel, or is wrapped in ExcludeSemantics because it is decorative. There is no third option — an unlabelled Icon is invisible to a screen reader.
  3. Read a11y state from MediaQuery at build time, never from app state. MediaQuery is an InheritedWidget with correct-by-construction invalidation; pushing it through a provider trades a compiler-guaranteed rebuild for a frame of staleness in the one area where being wrong is total failure.
  4. Never MediaQuery.withClampedTextScaling, never textScaleFactor. Both are clamping hacks that silently defeat the text-scale matrix while contrast and tap-target guidelines still pass green. Enforce with a source grep (see Anti-patterns).
  5. Never FittedBox, computed fontSize, or TextOverflow.ellipsis to make a label fit. They turn "doesn't fit at 200%" from a loud test failure into truncated or shrunk text on a device. Build flexible heights and let text wrap.
  6. Never encode state through color alone. Pair every stateful signal with at least one non-color channel — icon + text label + shape/pattern + position. Color-only state is invisible under invertColors, grayscale color-correction, and to every screen-reader user.
  7. Contrast is measured against the composited background in both themes — body/icon text ≥ 4.5:1 (AA 1.4.3), large/hero text ≥ 3:1 (AA) — large text never has to clear a higher bar than body. Never test against the nominal token color when a fill, gradient, or translucency sits between text and surface.
  8. Every interactive target is ≥ 44×44 logical pixels and single-tap. No long-press or precise-gesture-only affordances; hand-off and motor-impaired use demand a large forgiving target.
  9. Traversal order is authored with sortKey, never inherited from layout — visual position optimized for the thumb actively pessimises linear screen-reader/switch scanning.
  10. Honor boldText; keep haptics through reduced motion. Hardcoding fontWeight throws the setting away. Read disableAnimations from MediaQuery here, but resolve how motion collapses through design-system-structure's reduced-motion token + resolveMotion helper (that skill owns the mechanism) — an animation that still conveys meaning with no reduced-motion path is a bug.

Where a11y state comes from

Read platform accessibility state from context, app state from your notifier:

dart
@override
Widget build(BuildContext context) {
  final boldText = MediaQuery.boldTextOf(context);
  final highContrast = MediaQuery.highContrastOf(context);
  final reduceMotion = MediaQuery.disableAnimationsOf(context);
  // textScaler is deliberately NOT read and NOT clamped. Text scales itself.
  ...
}

MediaQuery.highContrastOf is iOS-only and always false on Android (AccessibilityFeatures.highContrast is documented "Only supported on iOS"). Read the flag opportunistically; never gate anything on it, and provide an in-app contrast/theme control as the mechanism that works everywhere. invertColors is a system compositing filter — never reimplement it; just never encode meaning in color alone, or an inverted screen changes what the UI says.

The tap target, wrong and right

dart
// WRONG — invisible to a screen reader, a clamping hack, a hidden overflow bug
class ItemTile extends StatelessWidget {
  const ItemTile({super.key, required this.item, required this.onTap});
  final Item item;
  final VoidCallback onTap;

  @override
  Widget build(BuildContext context) {
    return MediaQuery.withClampedTextScaling(   // (1) BANNED clamp
      maxScaleFactor: 1.3,
      child: GestureDetector(
        onTap: onTap,
        child: Column(
          children: [
            const Icon(Icons.star),             // (2) no semanticLabel
            Text(
              item.title,                        // (3) no Semantics button role
              style: const TextStyle(
                fontSize: 18,
                fontWeight: FontWeight.normal,   // (4) ignores boldText
              ),
              overflow: TextOverflow.ellipsis,   // (5) HIDES the overflow bug
            ),
          ],
        ),
      ),
    );
  }
}
dart
// RIGHT — semantics authored, text free to scale, target ≥ 44px, non-color state
class ItemTile extends StatelessWidget {
  const ItemTile({super.key, required this.item, required this.onToggle});
  final Item item;
  final void Function(ItemId id) onToggle; // resolve at tap from the stable id

  @override
  Widget build(BuildContext context) {
    final bold = MediaQuery.boldTextOf(context);
    final theme = Theme.of(context);
    final l10n = AppLocalizations.of(context); // non-null getter (i18n-rtl-l10n owns this contract)

    return Semantics(
      container: true,
      button: true,
      label: item.title,                       // display label, not a sentence
      // Non-color state: read out AND flip an icon (below), never color alone.
      value: item.isSelected ? l10n.selected : l10n.notSelected, // localized, not literal
      sortKey: OrdinalSortKey(item.priority.toDouble()), // priority, not layout
      child: GestureDetector(
        behavior: HitTestBehavior.opaque,       // whole cell is the target
        onTap: () => onToggle(item.id),         // stable id, no stale-closure capture
        child: ConstrainedBox(
          constraints: const BoxConstraints(minWidth: 44, minHeight: 44),
          child: ExcludeSemantics(              // face already announced above
            child: Row(
              children: [
                Icon(item.isSelected ? Icons.check_circle : Icons.circle_outlined),
                Text(
                  item.title,                   // no ellipsis, no FittedBox
                  style: theme.textTheme.bodyLarge!.copyWith(
                    fontWeight: bold ? FontWeight.bold : FontWeight.normal,
                  ),
                ),
              ],
            ),
          ),
        ),
      ),
    );
  }
}

The label is the display label, never a longer vocalization — nothing in the type system distinguishes the two Strings, so a scanning user hears the whole sentence on every step unless you keep them separate. Resolve the action from the immutable id at tap time; capturing a mutable field into the closure fires the stale value on a fast re-tap.

Text scale: the instinct is the bug

The reflex when a fixed layout overflows at 200% is to disable text scaling for stability. That reflex is the defect. Auto-shrinking — FittedBox, computed fontSize, maxLines + ellipsis — is the identical bug in disguise: the layout stays tidy and the user's setting stops working, silently, with every guideline still green.

  • Honor MediaQuery.textScalerOf(context) by not touching it. Flutter text widgets already scale.
  • textScaleFactor is deprecated in favour of TextScaler precisely to support Android's nonlinear scaling; its use is almost always a clamp.
  • Build intrinsic/flexible heights; let text wrap to as many lines as it needs. Hero and display sizes may cap their own growth and wrap to a second line rather than truncate.
  • Reflow horizontal rows (chip bars, toolbars) at accessibility text sizes instead of clipping.
  • Use FontFeature.tabularFigures() (or .monospacedDigit()-style tabular numerals) wherever numbers change or align so they don't reflow or jitter.
  • Let overflow scream in tests. A red/yellow overflow stripe in a widget test is the feedback loop; a truncated word on a device is a failure nobody reports.

Never state through color alone

A real state — selected, active, pending, complete, error, disabled — is knowable without color. Encode it redundantly across independent channels:

  • Icon / glyph whose shape differs per state (grayscale-legible).
  • Text label — the word for the state, routed through localization.
  • Shape / pattern / outline — a monotonic stripe, keyline, or notch.
  • Position — sort the item that needs attention to the top under a header.

Expose the state to assistive tech through Semantics — a value/hint, or a liveRegion for a transient announcement — never through a luminance step alone:

dart
Semantics(
  liveRegion: true, // announced when it changes, e.g. "Sync complete"
  label: statusLabel(status), // the word, not the color
  child: StatusBadge(status: status), // icon + label + shape, color is decoration
)

Pair visual feedback with HapticFeedback.selectionClick() so confirmation survives reduced motion and low vision. Guard any latched "active" visual with a timeout that force-clears it — a stuck state is both a visual lie and a stale semantic value.

Contrast against the composited background

Verify contrast against what the pixel actually composites to, in both light and dark, not against the nominal token:

  • Body text and meaningful icons/outlines ≥ 4.5:1 (AA 1.4.3); push to 7:1 (AAA 1.4.6) for body where you can reach it. Large/hero text ≥ 3:1 (AA) or 4.5:1 (AAA) — large text is never required to exceed body.
  • Never render a critical number or label on a gradient, translucent glass, or chip fill; put it on an opaque surface.
  • Any accent fill that carries text must be pre-paired with a foreground that clears 4.5:1 on that fill, re-verified in dark mode (dark values are hand-tuned, never auto-flipped).
  • Avoid the red/green trap: a neutral value is neutral ink, "success" is green + ✓ glyph + the word, danger-red appears only on destructive actions.
  • Verify the whole UI in a grayscale pass and deuteranopia / protanopia / tritanopia simulation.

Traversal order is a design decision

Layout optimized for thumb reach (highest-priority controls low and central) combined with Flutter's default row-major traversal makes the most important control the last thing a linear scanner reaches. Decouple traversal from position with one argument:

dart
Semantics(
  sortKey: OrdinalSortKey(item.priority.toDouble()), // authored from priority
  ... )

Assert sortKey order equals priority order (not layout order) in a test. Know the limit: Flutter publishes no Switch Access support statement and no API simulates scanning or group selection, so this is a regression guard on intent — real-device passes with TalkBack, VoiceOver, and Switch Access are the only conformance evidence. Touch, switch, and screen reader are three different channels: design for all three or state plainly which one was dropped.

Anti-patterns

  • MediaQuery.withClampedTextScaling / textScaleFactor — the one-line "fix" a future contributor reaches for when an overflow test goes red; it defeats the entire text-scale matrix silently. Ban with grep -rn "withClampedTextScaling\|textScaleFactor" lib/ in CI.
  • FittedBox / computed fontSize / ellipsis on a real label — same bug wearing a disguise; the setting stops working while the layout stays tidy.
  • Unlabelled Icon/Image — invisible to a screen reader; must be labelled or ExcludeSemantics.
  • Reading boldText/highContrast/disableAnimations from a provider — one frame stale, in the area where wrong is total failure.
  • Gating anything on highContrastOf — always false on Android; it silently disables your feature on the majority platform.
  • Color as the only channel for a state — dies under invertColors, grayscale mode, and screen readers.
  • Capturing a mutable field into an onTap closure — fires the stale value on a fast re-tap; resolve from the stable id at tap time.
  • A precise-gesture-only or long-press-only affordance — excludes motor-impaired and hand-off use; provide a ≥44px single tap.
  • containsSemantics(...) in tests — deprecated; use isSemantics(...).

Definition of done

For any widget with a tap target or state:

  1. Semantics node with button: true, a display-only label, and an authored sortKey.
  2. Every Icon/Image labelled or explicitly ExcludeSemantics.
  3. No clamp, no FittedBox, no ellipsis on a real label; renders at TextScaler.linear(2.0) without overflow.
  4. boldText honored; no hardcoded fontWeight on user-facing text.
  5. Every state carries a non-color channel; transient state uses liveRegion or value; feedback includes a haptic.
  6. Contrast verified against the composited background in both themes (≥4.5:1 body / ≥3:1 large) and in a grayscale + CVD pass.
  7. Interactive targets ≥ 44×44 and single-tap.
  8. A test asserting the semantics with isSemantics(...) and traversal against priority order.

Related skills

  • widget-golden-and-a11y-testing — the guideline matchers, golden lanes, RTL goldens, and honest a11y-conformance limits that verify this authoring.
  • i18n-rtl-l10n — state labels and announcements routed through gen-l10n/ARB, and Directional geometry for correct-by-construction RTL.
  • design-system-structure — token→theme layering, hand-authored ColorScheme, the no-raw-values gate that keeps contrast pairings honest, and the reduced-motion token + resolveMotion helper it owns (this skill reads the flag, that skill resolves the motion).
  • state-management-riverpod — the split both skills co-own: app/domain state flows through Riverpod notifiers, while platform/a11y flags are read from MediaQuery/BuildContext and never routed through a provider.
  • widget-composition — small const Views, dumb widgets, and dispose discipline that this semantics work sits on top of.
  • design-review-workflow — the end-of-build screenshot sweep (light/dark × RTL × largest text × reduce-motion) where a11y floors are always blocking.

References

Frequently asked questions

What does the Accessibility As Code AI skill do?

Enforces accessibility as a correctness property authored into each widget — Semantics(button/label) or ExcludeSemantics on every node, a11y state read from MediaQuery not app state, never MediaQuery.withClampedTextScaling / textScaleFactor / FittedBox / TextOverflow.ellipsis to fit a label, non-color redundant channels (icon+label+shape+text) for every state, contrast against composited backgrounds (4.5:1 body / 3:1 large), 44px single-tap targets, OrdinalSortKey traversal, and honoring boldText / reduce-motion. Use when adding a GestureDetector/InkWell or any tap target, adding an Icon or...

Why use Accessibility As Code on TypingMind?

Because you install it once and use it with any model. Accessibility As Code 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 Accessibility As Code in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/zakariaf/Flutter-Skills/tree/main/skills/accessibility-as-code. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Accessibility As Code?

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 Accessibility As Code?

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

Is the Accessibility As Code AI skill free?

Yes. It is published on GitHub by zakariaf under the MIT 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 👇