VGV AI Flutter Plugin

A Claude Code plugin that accelerates Flutter & Dart development with best-practices skills from Very Good Ventures.

Developed with 💙 by Very Good Ventures 🦄

Overview

VGV AI Flutter Plugin is a collection of contextual best-practices skills that Claude uses when helping you write Flutter and Dart code. Each skill provides opinionated, production-quality guidance covering architecture patterns, naming conventions, folder structures, code examples, testing strategies, and anti-patterns to avoid, so you get code that follows VGV standards out of the box.

Installation

One-line install from your terminal:

claude plugin marketplace add VeryGoodOpenSource/very-good-claude-code-marketplace && claude plugin install vgv-ai-flutter-plugin

Or inside an active Claude Code session, run these as two separate commands (the second only after the first completes):

1. Add the marketplace:

   /plugin marketplace add VeryGoodOpenSource/very-good-claude-code-marketplace
   

2. Install the plugin:

   /plugin install vgv-ai-flutter-plugin
   

For more details, see the Very Good Claude Marketplace.

Skills

Agents

This plugin ships subagents that Claude Code can dispatch as isolated, specialized reviewers. Unlike skills, agents are not invoked as slash commands — Claude dispatches them automatically, or you can ask Claude to run one by name (e.g. "review my changes with the flutter-reviewer agent").

Hooks

This plugin includes SessionStart, PreToolUse, and PostToolUse hooks that validate the Very Good CLI, guard against CLI bypass, and automatically run Dart analysis and formatting on .dart files.

Prerequisites

• Dart SDK — must be available on your PATH
• jq — used to parse the hook payload; hooks are skipped gracefully if jq is not installed

The Very Good CLI is resolved from PATH, falling back to $PUB_CACHE/bin (default ~/.pub-cache/bin) for hosts that launch hooks without an interactive shell's PATH. The installed very_good is a shim that execs dart, so when the binary is found but its version cannot be read the check is reported as inconclusive and the hook stands aside rather than reporting the CLI as missing.

Evals

Skill evals ask whether Claude routes to a skill and follows it. claude plugin eval runs each case with this plugin loaded and again with no plugin at all, so a grader that passes in both arms is measuring the model rather than the skill. Runs authenticate the same way your normal Claude Code sessions do, so they need no API key locally.

claude plugin eval . --scaffold

Run them locally before opening a PR that changes a skill. CI also runs them after a merge to main, scoped to the skills that changed, as an advisory signal rather than a gate — see evals/README.md.

See evals/README.md for the case format, the grader reference, prerequisites, and what these evals deliberately do not cover.

Usage

Skills activate automatically when Claude detects relevant context in your conversation. Simply ask Claude to help with a Flutter or Dart task, and the appropriate skill's guidance will be applied.

For example:

You: Create a new Bloc for user authentication with login and logout events.

Claude: (applies the Bloc skill — uses sealed classes for events and states, follows the Page/View separation pattern, generates blocTest() tests with mocktail mocks, and follows VGV naming conventions)

You can also invoke skills directly as slash commands:

/create-project
/animations
/accessibility
/bloc
/internationalization
/layered-architecture
/material-theming
/navigation
/static-security
/testing
/ui-package
/license-compliance
/dart-flutter-sdk-upgrade
/very-good-analysis-upgrade
/green-gate

What Each Skill Provides

Every skill includes:

• Core Standards — recommended conventions (e.g., mocktail over mockito, sealed classes for Bloc events)
• Architecture patterns — folder structures and layered architecture guidance
• Code examples — ready-to-adapt snippets following best practices
• Testing strategies — unit, widget, and integration testing patterns
• Common workflows — step-by-step guides for tasks like "adding a new feature" or "adding a new route"
• Anti-patterns — what to avoid and why

MCP Integration

This plugin includes a .mcp.json configuration that connects Claude Code to two MCP servers — the Dart and Flutter MCP server (dart mcp-server) and the Very Good CLI MCP server (very_good mcp). This gives Claude the ability to execute Dart, Flutter, and Very Good CLI actions directly, complementing the skills which provide architectural guidance and best practices.

Dart and Flutter MCP server (dart)

The Dart and Flutter MCP server ships with the Dart SDK and exposes core Dart/Flutter development actions to Claude.

Available MCP tools:

The server's cli feature category — dart_format, dart_fix, create_project, run_tests, and list_devices — is off by default, so a bare dart mcp-server advertises 13 tools with no formatter. .mcp.json therefore starts it as dart mcp-server --enable dart_format. The green-gate skill's format gate calls that tool; drop the flag and the gate has nothing to call. Test execution is deliberately left disabled — the test tool from the Very Good CLI MCP server is used instead.

Prerequisites:

• Dart SDK installed with dart on your PATH (the MCP server is provided by dart mcp-server)

Very Good CLI MCP server (very-good-cli)

The Very Good CLI MCP server exposes Very Good CLI commands to Claude.

Available MCP tools:

Prerequisites:

• Very Good CLI v1.5.0+ installed: dart pub global activate very_good_cli
• very_good must be on your PATH

How it works:

The .mcp.json file at the project root registers the dart and very-good-cli MCP servers using stdio transport. When Claude Code detects this configuration, it connects to both servers and gains access to the tools above. The skills continue to provide knowledge and best practices while the MCP tools handle execution.