agentleFS
evancaevanca/flutter-ai-rules/skills/flutter-best-practices/SKILL.md

flutter-best-practices

Skill669 starsChanged 3 months agoMITView on GitHub

Use when writing, reviewing, refactoring, or planning Flutter/Dart code — screens, features, project structure, state management, folders, widgets, cubits/blocs, repositories, services, or tests.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the reference file that matches the task.

    line 18
  2. Distinctive:

    Add layers only when they pay for themselves: Start with view-model → API client.

    line 52
  3. Very specific:

    Prefer intent-revealing widgets over Container: Use Padding, SizedBox, ColoredBox, DecoratedBox, Center — they are const-able and …

    line 60
  4. Very specific:

    Tests tell a story: Use expressive matchers (isEmpty, throwsA, isA, completion) and minimize dependencies — plain Text/SizedBox over …

    line 70
  5. Distinctive:

    Add unit tests for the cubit/view model logic before wiring UI details.

    line 89
  6. Distinctive:

    Check, in order of importance:

    line 93
  7. Distinctive:

    If more than ~2 teams or clearly separable domains are involved, read references/enterprise-scale.md and consider a Melos monorepo with one …

    line 109

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Flutter Best Practices

Standards for building maintainable Flutter apps, distilled from the official Flutter architecture guide and LeanCode's experience shipping 40+ Flutter projects (including a 30-developer banking app). Apply these when writing new code; when touching existing code, prefer consistency with the surrounding codebase and raise conflicts with these standards rather than silently rewriting.

How to use this skill

Read the reference file that matches the task. Read more than one when tasks overlap (e.g. a new feature touches both structure and coding style).

TaskRead
Design layers, decide where logic lives, MVVM, repositories/servicesreferences/architecture-layers.md
Create/organize a feature, folder structure, state management wiringreferences/feature-structure.md
Write or review Dart/Flutter code, widgets, tests, lintsreferences/dart-coding-practices.md
Multi-team/multi-package apps, monorepo, navigation, localization, API contracts, E2E testsreferences/enterprise-scale.md
App localization setup, reusable UI package string ownership, language picker visibilityreferences/localization-package-boundaries.md

For a quick task (small widget fix, one-line review comment), the core rules below may be enough on their own.

Core rules (always apply)

Architecture

  1. Separate UI from data. Two broad layers: UI (views + view models / cubits) and Data (repositories + services). Dependencies point one way: View → ViewModel → Repository → Service. Lower layers never import upper layers. Repositories never depend on each other.
  2. Views hold no business logic. Widgets may contain show/hide conditionals, animation, layout, and simple routing logic — nothing that transforms or decides about data. All data logic lives in the view model (or cubit/bloc), which has no access to BuildContext.
  3. Organize by feature, not by layer. Everything a feature needs — state management, widgets, models — lives under one feature directory. Don't create top-level blocs/, widgets/, models/ buckets that scatter a feature across the tree.
  4. State is immutable and explicit. Model UI state as a sealed/union type (initial / inProgress / failure / ready) so every case is handled exhaustively. One-off effects (snackbars, navigation) are events, not state.
  5. Add layers only when they pay for themselves. Start with view-model → API client. Introduce a repository when you need caching, offline, or merging sources. Introduce a use case only when logic merges multiple repositories, is genuinely complex, or is reused by several view models.

Coding

  1. Prefer intent-revealing widgets over Container. Use Padding, SizedBox, ColoredBox, DecoratedBox, Center — they are const-able and self-describing. Container is fine only when combining several properties at once.
  2. Use modern Dart. Pattern matching (if (x case final v?)), switch expressions with exhaustiveness, records and destructuring, collection if/for/spreads instead of .add() loops, expression bodies for pass-through async functions (no redundant async/await).
  3. Prefix sliver-returning widgets with Sliver so misuse in the wrong scroll context is caught at a glance.
  4. Tests tell a story. Use expressive matchers (isEmpty, throwsA, isA, completion) and minimize dependencies — plain Text/SizedBox over design-system widgets in test fixtures. Test cubits/view models in isolation from the widget tree.
  5. Every // ignore: gets a reason on the same or preceding line. Log errors with dedicated error/stackTrace parameters, never string interpolation.

Workflow checklists

Adding a new feature/screen

  1. Read references/feature-structure.md and mirror the existing project's conventions for the feature directory.
  2. Define the state as a union type first; then the cubit/view model; then the widgets. Constructor-inject dependencies; scope them to the feature's widget subtree.
  3. Data comes in through a repository or typed API client — never fetched inside a widget.
  4. Add unit tests for the cubit/view model logic before wiring UI details.

Reviewing Flutter code

Check, in order of importance:

  1. Logic in the right layer (rule 1–2)? Any BuildContext in business logic?
  2. State modeled as immutable union types, all cases handled?
  3. Feature self-contained, or does it reach into another feature's internals?
  4. Widget choices (rule 6), modern Dart (rule 7), sliver naming (rule 8)?
  5. Tests present for logic, readable, minimal dependencies?
  6. Unexplained // ignore:, string-interpolated error logs, deprecated patterns still spreading?

Starting a new project

  1. Read references/architecture-layers.md for the layer blueprint and references/feature-structure.md for the folder skeleton.
  2. If more than ~2 teams or clearly separable domains are involved, read references/enterprise-scale.md and consider a Melos monorepo with one package per domain from day one.
  3. Set up strict lints early (leancode_lint or equivalent + custom rules) — retrofitting is far more expensive.

Package palette

Defaults that these standards assume (swap for project-local equivalents when the codebase already uses something else):

  • State: bloc (Cubit) + freezed for union-type states; bloc_presentation for one-off UI events
  • DI: provider scoped to widget subtrees (accepting its lack of compile-time safety as the lesser evil)
  • Boilerplate reduction: flutter_hooks
  • Monorepo: melos
  • Localization: flutter_localizations + intl with .arb files
  • Lints: leancode_lint
  • E2E/UI tests: patrol

The official Flutter guide is state-management-agnostic (MVVM with ChangeNotifier works too); what matters is the layer separation, not the package. See the reference files for rationale and trade-offs.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.