This repository offers a suite of specialized AI agents designed to enhance Rails development, catering to distinct architectural philosophies and promoting Test-Driven Development (TDD). It targets Rails developers seeking AI-driven assistance with deep context, aiming to accelerate development, enforce best practices, and provide architectural flexibility.

How It Works

The suite presents two primary architectural philosophies: the "37signals Style," which favors rich models and a CRUD-centric approach, and "Standard Rails," emphasizing service objects, query objects, and presenters for clear separation of concerns. A "Skills Library" complements these by offering reusable knowledge modules on specific Rails patterns and technologies. Agents are categorized under these philosophies, allowing developers to select or combine styles, and are engineered to support TDD workflows, including generating initial failing tests and assisting with refactoring.

Quick Start & Requirements

• Installation: Specific installation commands are not detailed, but usage is implied within a Rails project context.
• Prerequisites: Ruby 3.3+ is required. Agents target Rails 8.x (edge) for the 37signals style and Rails 7.x for the Standard style. Databases like PostgreSQL or SQLite are supported. Hotwire (Turbo + Stimulus) is a core dependency. The 37signals agents utilize Solid Queue/Cache/Cable, while Standard agents integrate with ViewComponent, Tailwind CSS, Pundit, and RSpec + FactoryBot.
• Documentation: References are made to CLAUDE_CODE_SETUP_PLAN.md and "The Complete Guide to Claude Code V4" for detailed Claude Code setup instructions.

Highlighted Details

• Provides two distinct Rails architectural paradigms: "37signals Style" (rich models, concerns, CRUD-everything) and "Standard Rails" (service objects, query objects, presenters).
• Includes a "Skills Library" offering reusable knowledge modules covering specific Rails patterns and technologies.
• Features integration with Claude Code, complete with setup plans for security hooks, custom commands, and MCP servers.
• Agents are designed to adhere to TDD principles, with dedicated agents for writing initial failing tests (@tdd_red_agent) and performing code refactoring (@tdd_refactoring_agent).

Maintenance & Community

The provided documentation does not specify maintainers, community channels (e.g., Discord/Slack), or a project roadmap. It notes that the agents were built using insights from GitHub's analysis of agent.md files and 37signals' Fizzy codebase.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license permits unrestricted use, modification, and distribution, making it compatible with commercial and closed-source projects.

Limitations & Caveats

Some agents are designed for "Rails 8.x (edge)," indicating potential instability or breaking changes associated with bleeding-edge Rails versions. The project's reliance on AI for development implies inherent limitations tied to the current capabilities and potential inaccuracies of AI-generated code or suggestions.