Flutter Agent Plugins

The dart_skills_lint tool is designed to validate Agent Skill specifications (typically SKILL.md files). It follows a layered architecture that separates CLI handling, configuration parsing, and validation logic.

Key Components

• CLI Entry Point: Handles argument parsing (e.g., --skills-directory), workspace discovery (searching .agents/skills or .claude/skills), and log management.
• Configuration Parser: Loads custom settings from dart_skills_lint.yaml to manage directory-specific rule overrides and severity defaults.
• Validation Engine: The core component that scans SKILL.md files using regular expressions to extract Frontmatter and validates directory structures, field constraints, and relative links.
• Predefined Rules: A collection of global definitions for standard checks using CheckType.
• Core Data Models: Defines the structure for ValidationError, CheckType, and IgnoreEntry (used for suppressing errors via JSON).

Use test_process and test_descriptor to write high-fidelity integration tests that validate filesystem mutations and process execution.

Testing Workflow:

1. Define expected filesystem states using test_descriptor (d.dir, d.file).
2. Create the mock filesystem with await d.Descriptor.create().
3. Spawn the CLI process using TestProcess.start('dart', ['run', 'bin/cli.dart', ...args]).
4. Validate stdout and stderr using StreamQueue matchers (e.g., emitsThrough, emits).
5. Assert the exit code using await process.shouldExit(0).
6. Validate filesystem changes using await d.Descriptor.validate().

import 'package:test/test.dart';
import 'package:test_process/test_process.dart';
import 'package:test_descriptor/test_descriptor.dart' as d;

void main() {
  test('CLI test example', () async {
    await d.dir('project', [d.file('config.json', '{}')]).create();

    final process = await TestProcess.start(
      'dart',
      ['run', 'bin/cli.dart', 'process', '--path', '${d.sandbox}/project']
    );

    await expectLater(process.stdout, emitsThrough('Processing complete.'));
    await process.shouldExit(0);
    await d.dir('project', [d.file('config.json', '{}')]).validate();
  });
}

When using package:ffigen, adhere to these constraints:

• Never write manual bindings: If native headers exist, always use FfiGenerator instead of manual DynamicLibrary.lookup or @Native functions.
• Standard Locations: Place the generator at tool/ffigen.dart. Place third-party headers in third_party/ or src/ at the package root.
• Output Directory: Generated files for third-party libraries must be placed under lib/src/third_party/.
• File Extensions: The primary generated FFI bindings file must use the .g.dart extension.
• Preamble: Always include a preamble in the Output class that specifies the license, includes the native copyright header, and contains an automatic generation warning (e.g., // Generated by package:ffigen. Do not edit manually.).
• Tree Shaking: To support native asset tree shaking, set recordUse: (_) => true in Functions and specify a recordUseMapping target in Output (e.g., lib/src/third_party/sqlite3.record_use_mapping.g.dart).

Follow these steps sequentially when adding a new feature:

1. Define Domain Models: Create immutable data classes (e.g., using freezed).
2. Implement Services: Create classes to handle external API communication.
3. Implement Repositories: Create the Repository to consume Services and return Domain Models.
4. Apply Conditional Logic (Domain Layer): Create a Use Case class only if the feature requires complex data transformation or cross-repository logic.
5. Implement the ViewModel: Create a ChangeNotifier ViewModel. Inject Repositories/Use Cases and expose immutable state.
6. Implement the View: Create the UI widget. Use ListenableBuilder or AnimatedBuilder to listen to the ViewModel.
7. Inject Dependencies: Register the Service, Repository, and ViewModel in your DI container (e.g., provider or get_it).
8. Run Validator: Execute unit tests for the ViewModel and Repository.