- Updated: March 31, 2026
- 4 min read
Claude Code Rails Skills – Accelerate Your Rails Development
**Summary – Claude Code “Rails Skills” (jr‑rails‑classic, jr‑rails‑new, jr‑rails‑phlex)**
—
## 1. What the article is about
The piece introduces a **set of three Claude Code “skills”** that teach an AI assistant how to write Ruby on Rails applications the way a seasoned 37signals/HEY team would.
* They are **opinionated, production‑ready** and aim to keep the AI from reinventing the wheel, which reduces bugs, maintenance cost and development time.
* The skills can be installed **all at once** via the CLI (`npx skills add julianrubisch/skills`) or manually by dropping the skill folders into `.claude/skills/` in a project.
—
## 2. The three skills – purpose & usage
| Skill | What it does | How you invoke it | Core features | What you gain |
|——-|————–|——————-|—————|—————|
| **jr‑rails‑classic** | Teaches the AI to code Rails **the way 37signals does** (the “classic” Basecamp/HEY stack). | `plugin install jr-rails-classic@julianrubisch-skills` (or auto‑picked after install). | • Rich models, thin controllers
• CRUD controllers built with concerns
• Minitest + fixtures
• **No service objects** (keeps logic in models/presenters) | • Less custom code → fewer bugs, lower maintenance
• Consistent, battle‑tested conventions |
| **jr‑rails‑new** | Walks the AI through **bootstrapping a brand‑new Rails app** with a proven stack. | Same install command; then the AI runs an **interactive interview** before scaffolding. | • Full post‑scaffold configuration
• Choice of **Phlex or ERB** for views
• Integrated gems: **Pundit, Solid Queue, AASM**
• Production‑ready defaults (Hotwire, Turbo, etc.) | • New projects start on solid ground → no costly re‑work later
• Immediate readiness for scaling, hand‑off, or team growth |
| **jr‑rails‑phlex** | Guides the AI in building **UI components with Phlex** (or similar component‑based view layer). | Same install; AI calls the skill when generating UI. | • Component hierarchy & slots
• Stimulus & Turbo integration
• Custom elements, helpers, fragment‑caching patterns
• Uses Rails scaffold generator as a base | • Consistent, fast, testable UI
• One‑place updates → no “whack‑a‑mole” UI bugs |
**Installation shortcut** – Running `npx skills add julianrubisch/skills` installs **all three** at once, so you can start using any of them immediately.
—
## 3. Why the skills matter – “With vs. Without” comparison
| Area | Without the skills | With the skills |
|——|——————-|—————–|
| **Login security** | AI writes a 200+‑line custom auth system → hidden security holes. | Uses Rails’ built‑in, battle‑tested authentication (few lines, proven). |
| **Code bloat** | AI adds extra gems / custom solutions for problems Rails already solves → more dependencies, more breakage. | AI leans on native Rails features → leaner codebase, fewer surprises. |
| **State / workflow tracking** | Logic scattered, no audit trail → hard debugging. | Explicit, guarded, timestamped state changes (e.g., AASM) → easy traceability. |
| **UI consistency** | Repeated copy‑paste of view snippets → divergent pages, maintenance nightmare. | Reusable Phlex components → single‑source updates, uniform UI. |
Overall, the skills **reduce the amount of code the AI generates**, keep the code **aligned with proven Rails conventions**, and therefore **lower cost, risk, and technical debt**.
—
## 4. What’s inside the skill packages
* **Deep reference material** – architecture guides, design patterns (decorator, presenter, form object, etc.), refactoring recipes, anti‑pattern catalog.
* **Domain guides** – common building blocks (auth, notifications, background jobs, state machines).
* **Hotwire integration patterns** – Stimulus, Turbo, real‑time UI.
* **Testing strategy** – Minitest fixtures, best‑practice test layout.
* **Security & authorization** – Pundit policies, Noticed notifications.
* **Background processing** – Solid Queue, AASM state machines, Rails.event handling.
* **Deployment** – Kamal + Thruster scripts.
The skill set is **opinionated by design**: every tool and pattern is a deliberate choice, eliminating decision fatigue and endless tooling debates.
—
## 5. Overall benefits
1. **Higher quality output** – AI follows a vetted, production‑grade Rails stack.
2. **Speed & efficiency** – Less custom code to write, test, and maintain.
3. **Consistency** – Uniform architecture, UI components, and conventions across the whole app.
4. **Scalability & hand‑off readiness** – Apps start with production‑ready gems and patterns, making future scaling or team changes painless.
5. **Reduced security risk** – Leverages Rails’ built‑in, battle‑tested mechanisms instead of reinventing them.
6. **Focused AI** – The agent pulls only the relevant reference material for the task, staying lightweight and avoiding “information overload.”
—
### TL;DR
The **jr‑rails‑classic**, **jr‑rails‑new**, and **jr‑rails‑phlex** skills are a bundled, opinionated toolkit for Claude Code that teaches the AI to write Rails applications exactly the way 37signals does: minimal custom code, strong conventions, production‑ready defaults, and component‑based UI. Installing them (via CLI or manual copy) instantly equips the AI to generate cleaner, safer, and more maintainable Rails code, saving time, money, and headaches for developers.
Andrii Bidochko
CTO UBOS
Andrii Bidochko is an AI entrepreneur and researcher focused on AI agents, reinforcement learning, and autonomous systems. He writes about the technologies shaping the future of machine intelligence, from frontier models and agent architectures to real-world AI applications.