Ui States And Feedback logo

Ui States And Feedback

Community
zakariaf
ui-states-and-feedback

Enforces the non-happy paths every screen owes the user — loading/empty/error/content resolved in ONE switch over an AsyncValue or sealed view state, never an isLoading+error+data bool triple; a delayed skeleton shaped like the result instead of a spinner that flashes; a refresh that keeps stale content on screen; "nothing yet" distinguished from "nothing matches this filter", each with its own action; error text from a typed Failure code (never e.toString()) plus a retry that invalidates the provider; the surface ladder inline, snackbar, banner, dialog, with a modal earned only by a decision that must resolve now; soft-delete + Undo over a confirmation; showDialog's null dismissal handled as its own outcome; ScaffoldMessenger captured before an await; and announcements for eye-only changes. Use when writing loading, empty, error, or offline UI, calling AsyncValue.when, showDialog, showModalBottomSheet, showSnackBar or showMaterialBanner, or adding a confirmation, Undo, or retry.

Overview

Publisherzakariaf
RepositoryFlutter-Skills
Skill nameui-states-and-feedback
Stars
64
Forks
15
Bundled files
1
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.

  • 1 bundled files

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

  • Open source

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

Installation

Install the Ui States And Feedback 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/ui-states-and-feedback
  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/ui-states-and-feedback .claude/skills/ui-states-and-feedback
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Ui States And Feedback 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 Ui States And Feedback 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 Ui States And Feedback 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 states and feedback

Most of a screen's code is the happy path; most of its defects are not. This skill governs everything a screen shows when there is no data yet, no data at all, no data matching the filter, a failure, or a write that just returned — and which surface each of those is allowed to use. The first-frame/cold-start path belongs to app-startup-and-bootstrap; the error values belong to error-handling-typed-results. This skill is only about how they are rendered and announced.

Non-negotiable rules

  1. One state, resolved in one place. A screen renders exactly one of loading / empty / error / content, chosen by a single switch over an AsyncValue or a sealed view state. Never a triple of isLoading, error, items booleans. WHY: separate flags make impossible states representable — loading and error, empty and loading — and every such combination eventually renders.
  2. The ViewModel decides the state; the View renders it. No FutureBuilder or StreamBuilder wrapped around a repository call in the middle of a widget tree. WHY: state scattered across the tree cannot be tested without pumping widgets, and rebuilds re-fire the future. See state-management-riverpod, widget-composition.
  3. Empty, filtered-empty, and error are three different screens. "You haven't added anything yet" (offer the primary action), "Nothing matches this filter" (offer to clear it), and "We couldn't load this" (offer retry) answer three different user questions. WHY: showing "no items" for a failed load teaches the user their data is gone.
  4. Never render an exception. No e.toString(), no stack trace, no failure class name in the UI. Map the typed Failure.code to a localized message via ARB. WHY: a raw exception is untranslatable, unactionable, and often leaks paths or ids.
  5. Every error state offers a next action. Retry (ref.invalidate(provider)), go back, or change the input — and retry must preserve scroll position, filters, and entered text. WHY: a dead-end error screen turns a transient failure into an uninstall.
  6. A refresh never blanks the screen. While reloading, keep the previous content and show a subtle indicator (AsyncValue.when's skipLoadingOnRefresh default, or valueOrNull + isRefreshing). WHY: replacing a list with a spinner loses the user's place and reads as data loss.
  7. Loading is a delayed skeleton shaped like the result — not a centered spinner. Suppress it for roughly the first 150–300 ms so a fast load does not flash, and match the final layout so nothing jumps when content arrives. WHY: a flash and a reflow are two separate perceived defects; both are free to avoid. (Every duration in this skill is a behavioral threshold, not a value to hardcode — the numbers themselves belong to the duration tokens in design-system-structure.)
  8. An in-flight action shows progress on its own control, not over the app. Disable the button, show progress in it, keep the rest interactive. A full-screen modal barrier is earned only when continuing would corrupt state. WHY: a blocking barrier over a 200 ms write is the most common self-inflicted "the app froze".
  9. Pick the surface by the ladder: inline < snackbar < banner < dialog. Take the lowest one that works (table below). A modal is earned only by a decision that must resolve before the user can continue. WHY: every step up the ladder takes control away from the user, and dialogs are the only one that also blocks.
  10. Information the user must act on never lives only in a snackbar. It auto-dismisses, it can be missed entirely, and a screen reader may never reach it. Use a banner or an inline state for anything that still matters in ten seconds.
  11. Prefer soft delete + Undo over a confirmation dialog. Confirm only what is genuinely irreversible, and then name the consequence ("Delete 12 notes permanently") rather than asking "Are you sure?", with barrierDismissible: false. WHY: Undo is faster for the 99% who meant it and safer for the 1% who did not. See error-handling-typed-results for the soft-delete write path.
  12. showDialog/showModalBottomSheet return Future<T?> — null means dismissed, and it is its own outcome. Switch over a sealed result and handle dismissal explicitly; never ?? true. WHY: ?? true silently converts a tap outside the dialog into a confirmed deletion.
  13. Capture ScaffoldMessenger/Navigator before the await, guard mounted after. Then show the snackbar. WHY: the BuildContext may be gone when the write returns — this is the exact use_build_context_synchronously hole (async-safety).
  14. Announce what only the eye can see. A result that appears (saved, deleted, retried, filtered) is announced with liveRegion: true or SemanticsService.announce, and auto-dismissal is lengthened or removed when MediaQuery.accessibleNavigationOf(context) is true. WHY: a 4-second snackbar is invisible to a screen-reader user mid-sentence. Announcements are best-effort per platform — never the only channel for something important.

The four states — one resolution point

dart
class NotesScreen extends ConsumerWidget {
  const NotesScreen({super.key});

  @override
  Widget build(BuildContext context, WidgetRef ref) {
    final l10n = AppLocalizations.of(context);
    final notes = ref.watch(notesProvider);

    // ONE switch. skipLoadingOnRefresh (default true) keeps the list on screen
    // while it reloads — the refresh path never falls into `loading`.
    return notes.when(
      loading: () => const NotesSkeleton(),          // shaped like the list, delayed
      error: (error, _) => ErrorStateView(
        // A typed Failure's stable code → a localized string. Never e.toString().
        message: l10n.failureMessage(error is Failure ? error.code : 'error.unknown'),
        onRetry: () => ref.invalidate(notesProvider),
      ),
      data: (notes) => notes.isEmpty
          ? EmptyStateView(                          // "nothing yet" ≠ "no matches"
              title: l10n.notesEmptyTitle,
              action: l10n.notesEmptyCreate,
              onAction: () => ref.read(notesProvider.notifier).createDraft(),
            )
          : NotesListView(notes: notes),
    );
  }
}

A screen with filters resolves filtered-empty before empty: the ViewModel knows whether a filter is active, so the View never has to guess why the list is short. Complete file, including the delayed skeleton and the partial-failure row state: examples/async_state_view.dart.

The surface ladder

SurfaceLifetimeUse it forNever use it for
Inline (in the field/region it concerns)Until resolvedValidation, a per-item failure, an empty regionA global condition
SnackbarSeconds, auto-dismissA completed action + its Undo; a non-critical, already-recovered failureAnything the user must act on, or anything they'd want to re-read
Banner (showMaterialBanner)Until dismissedAn app-level condition that persists: offline, expired data, a degraded modeA transient confirmation
Dialog / modal sheetBlocks until resolvedA decision that must resolve now, an irreversible confirmationShowing information, reporting a result, or anything you could render inline

A per-item failure inside a list is an inline row state, not a screen state — one row that failed to sync must not replace the other forty-nine.

Feedback after a write

dart
Future<void> _delete(BuildContext context, WidgetRef ref, NoteId id) async {
  // Capture BEFORE the await — the context may be gone when the write returns.
  final messenger = ScaffoldMessenger.of(context);
  final l10n = AppLocalizations.of(context);

  final result = await ref.read(notesProvider.notifier).softDelete(id);
  if (!context.mounted) return;

  switch (result) {
    case Ok():
      messenger.clearSnackBars(); // don't queue behind an older one
      messenger.showSnackBar(SnackBar(
        content: Text(l10n.noteDeleted),
        // Undo instead of a pre-emptive "Are you sure?" dialog.
        action: SnackBarAction(
          label: l10n.undo,
          onPressed: () => ref.read(notesProvider.notifier).restore(id),
        ),
        // Auto-dismiss is a barrier for screen-reader and switch users.
        duration: MediaQuery.accessibleNavigationOf(context)
            ? const Duration(seconds: 30)
            : const Duration(seconds: 6),
      ));
    case Err(:final failure):
      // A failure the user must act on is NOT a snackbar.
      messenger.showMaterialBanner(deleteFailedBanner(l10n, failure.code));
  }
}

SnackBarAction is the Undo affordance for a soft delete: the row is flagged, not destroyed, and the hard delete happens later (error-handling-typed-results). An Undo that cannot actually restore the record is worse than no Undo.

Confirming something irreversible

dart
// Sealed outcome — dismissal is a real case, not a missing value.
sealed class ConfirmOutcome {}
final class ConfirmAccepted implements ConfirmOutcome {}
final class ConfirmDeclined implements ConfirmOutcome {}
final class ConfirmDismissed implements ConfirmOutcome {} // tapped outside / back

final outcome = await showDialog<ConfirmOutcome>(
      context: context,
      barrierDismissible: false, // a decision that must resolve does not evaporate
      builder: (_) => PermanentDeleteDialog(count: selected.length),
    ) ??
    ConfirmDismissed(); // null == dismissed — NEVER `?? ConfirmAccepted()`

The dialog's title names the consequence and the count; the confirming button carries the verb ("Delete permanently"), never "OK". Destructive verbs and the count are what a user reads before tapping — "Are you sure?" tells them nothing.

Anti-patterns

  • isLoading + errorMessage + items as three fields — impossible states become representable and eventually render.
  • FutureBuilder over a repository call inside build() — refires on every rebuild and cannot be tested without a widget pump.
  • A centered spinner replacing loaded content on refresh — loses scroll position and reads as data loss.
  • A spinner with no delay — flashes on every fast load; the flash is the defect.
  • A skeleton that does not match the final layout — content lands and everything jumps, which reads as jank.
  • "No items" shown for a failed load — tells the user their data is gone.
  • Text('${e}') or a stack trace in the UI — untranslatable, unactionable, leaky.
  • An error state with no action — a transient failure becomes a dead end.
  • A dialog to report a result ("Saved!") — blocks the user to say something a snackbar or nothing at all could say.
  • A critical warning in a snackbar — auto-dismisses, may be missed, may never reach a screen reader.
  • ?? true on a dialog's result — a tap outside becomes a confirmed destruction.
  • ScaffoldMessenger.of(context) after an await with no mounted guard — throws or targets a dead tree (async-safety).
  • A full-screen blocking barrier for a fast local write — the app looks frozen.
  • A confirmation dialog on every delete instead of Undo — punishes the 99% who meant it and does not actually protect the 1%.

Definition of done

  • The screen resolves exactly one of loading / empty / error / content in a single switch/when; no bool triple; no FutureBuilder over a repository call.
  • Empty, filtered-empty, and error are distinct views; empty carries the primary action and filtered-empty offers to clear the filter.
  • Error text comes from a typed Failure.code mapped through ARB; no exception, class name, or stack trace is rendered.
  • Every error state has a retry/next action that preserves scroll, filters, and input.
  • Refresh keeps previous content on screen; loading is a delayed skeleton matching the final layout.
  • In-flight actions show progress on their own control; no app-wide barrier without a stated reason.
  • Each message uses the lowest adequate surface; nothing actionable lives only in a snackbar; per-item failures render inline in their row.
  • Destructive actions use soft delete + Undo, or a barrierDismissible: false dialog that names the consequence and count.
  • Dialog dismissal is handled as its own outcome; no ?? <affirmative>.
  • ScaffoldMessenger/Navigator captured before every await; mounted guarded after; snackbars cleared before showing a new one.
  • State changes are announced (liveRegion/SemanticsService.announce), and auto-dismiss is extended under accessibleNavigation.
  • Widget tests cover all four states plus dismissal and retry.

Related skills

  • state-management-riverpod — where the state is decided, and AsyncValue ownership.
  • error-handling-typed-results — the Result/Failure values rendered here, and the soft-delete/Undo write path.
  • async-safety — capturing context before an await and guarding mounted after.
  • widget-composition — dumb Views, extracted const state widgets, no logic in build().
  • accessibility-as-code — semantics, live regions, and the a11y flags read from MediaQuery.
  • motion-and-haptics — the sensory half of feedback: what the transition into each of these states animates, the one haptic per committed action, and the reduced-motion path.
  • i18n-rtl-l10n — every message in ARB, including failure-code messages.
  • app-startup-and-bootstrap — the first-frame/splash path this skill does not cover.
  • widget-golden-and-a11y-testing — pumping and asserting each of the four states.
  • design-system-structure — where the skeleton, banner, and empty-state components live.

References

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 Ui States And Feedback AI skill do?

Enforces the non-happy paths every screen owes the user — loading/empty/error/content resolved in ONE switch over an AsyncValue or sealed view state, never an isLoading+error+data bool triple; a delayed skeleton shaped like the result instead of a spinner that flashes; a refresh that keeps stale content on screen; "nothing yet" distinguished from "nothing matches this filter", each with its own action; error text from a typed Failure code (never e.toString()) plus a retry that invalidates the provider; the surface ladder inline, snackbar, banner, dialog, with a modal earned only by a decisi...

Why use Ui States And Feedback on TypingMind?

Because you install it once and use it with any model. Ui States And Feedback 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 Ui States And Feedback in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/zakariaf/Flutter-Skills/tree/main/skills/ui-states-and-feedback. 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 Ui States And Feedback?

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 Ui States And Feedback?

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

Is the Ui States And Feedback 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 👇