What does this skill do, and when should you use it?
This is an Agent Skill for Rails 8.x production codebases, covering models, controllers, routes, Hotwire, background jobs, APIs, performance, security, and testing. Its core principle is to inspect existing code first and follow local conventions, making migration-style changes only when the user requests them. The skill uses an on-demand reference-loading structure so only the sections relevant to a task are read. It is maintained by the Ethos Link team and released under the MIT license.
The skill instructs the agent to read the Gemfile, config/application.rb, config/routes.rb, environment configs, representative models and controllers, test directories, and deploy entrypoints before proposing changes, recording existing patterns like test framework, authentication style, frontend stack, and queue backend. It includes a queue-adapter detection procedure (checking queue_adapter config, then Gemfile, then worker runtime commands) to avoid wrongly assuming Solid Queue. It routes task types to 15 reference documents (e.g., models-and-data, security-checklist, hotwire-turbo-stimulus). For implementation tasks it outputs migrations, code, tests, and risk notes; for review tasks it prioritizes correctness, security, performance, and test gaps. The repo ships a scripts/validate_skill.rb validation script.
- A Rails team wants an AI assistant to add features (e.g., a billing webhook endpoint) to an existing production codebase while matching existing conventions and tests
- Reviewing a Rails 8 pull request for production readiness, focusing on naming, data integrity, security, and missing tests
- A GoodJob-based app needs a new background job without switching queue backends
- A greenfield Rails 8.1 project wants to follow built-in authentication generators and framework-native REST resource routing
- A team wants to establish a domain terminology document (docs/domain-terms.md) and naming rules, such as distinguishing local ids from external eids
How do you install this skill?
- The 16 references and validate_skill.rb cited by SKILL.md are not included in the review evidence; their content quality and security guidance are unverified.
- Publisher Ethos Link is not registry-verified, and the README contains visible product promotion — distinguish guidance from marketing.
- This is a static, non-executed review with low confidence; no scripts were run and skill behavior was not reproduced.
- No changelog or explicit update path; version-sensitive Rails 8.1 guidance may become stale as the framework evolves.
- Local filesystem
Optional: Ruby to run scripts/validate_skill.rbOptional: npx skills CLI for install
Install via the skills CLI: npx skills add [email protected]:ethos-link/rails-conventions.git --skill rails-conventions; update an existing install with npx skills update rails-conventions. Before publishing changes, validate with ruby scripts/validate_skill.rb. The repo does not document where SKILL.md should be placed for each specific runtime.
git clone --depth 1 https://github.com/ethos-link/rails-conventions.git ~/.claude/skills/rails-conventionsGenerated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Implement a new billing webhook endpoint in this Rails app, matching existing conventions and tests.
- Review this Rails 8 pull request for production readiness, focusing on naming, data integrity, security, and missing tests.
Place the repo in a runtime that supports the Agent Skills SKILL.md convention, then trigger it with Rails-related requests. Example prompts from the README: "Implement a new billing webhook endpoint in this Rails app, matching existing conventions and tests." or "Review this Rails 8 pull request for production readiness, focusing on naming, data integrity, security, and missing tests." The skill automatically loads the relevant references per task.
What are this skill's strengths and limitations?
- Emphasizes reading existing code before acting, preventing AI-imposed architecture that clashes with the project
- Built-in queue backend detection; never assumes Rails 8 equals Solid Queue and writes backend-agnostic ApplicationJob code
- References are routed by task, reducing irrelevant context consumption
- Explicit output contract and review priorities make results predictable
- Ships a validation script and forward-test prompts in the README to verify triggering behavior
- Targets Rails 8.x only; unsuitable for older Rails versions
- README cites Minitest and Minitest::Mock support; RSpec coverage is not mentioned
- No GitHub topics listed, and no public test suite or benchmarks demonstrating effectiveness
- The 37signals-style profile is an optional taste guide, not a fit for teams seeking it as the primary draw
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Rails Convention Engineer this page | 52 · Use with care | ★ 17 | 1mo ago | MIT |
| Loki Mode Autonomous SDLC Skill | 53 · Use with care | ★ 1.1k | 16d ago | NOASSERTION |
| Swift Concurrency Pro | 65 · Recommended | ★ 572 | 5mo ago | MIT |
| cmux Architecture Skill | 63 · Recommended | ★ 28k | 11d ago | NOASSERTION |
| Tool Rename Compatibility Check ✓ Microsoft · Official | 63 · Recommended | ★ 194k | 11d ago | MIT |
The README explicitly notes the skill is heavily inspired by the unofficial 37signals coding style guide (github.com/marckohlbrugge/unofficial-37signals-coding-style-guide), positioning it as taste guidance rather than authority; the source names no other competitors.
How did FollowSkills review this skill?
The skill is purely advisory: it executes no commands, requests no special permissions, and has no external side effects or exfiltration paths. It mandates codebase inspection first, explicit tradeoff reporting, and API verification against official docs — low-risk by design. Deducted because the 16 referenced files are not present in the evidence, so security guidance (auth patterns, security checklist) cannot be statically verified.
Instructions are internally consistent: task-routing tables match reference routing, queue-adapter detection is ordered explicitly (config → Gemfile → runtime), and tradeoffs must be reported, giving a controlled-failure mindset. Deducted because all references and the validate_skill.rb script are absent from evidence; key paths cannot be cross-checked, and under static review no execution evidence exists — failure-feedback quality is inferred only.
Trigger conditions in the description are concrete (Rails 8, Active Record, GoodJob, Solid Queue, Turbo/Stimulus, etc.), and README provides three forward-test prompts for trigger validation. Deducted because non-fit boundaries (non-Rails projects, older Rails versions) are only implicit and no explicit out-of-scope statement exists; no Chinese-language support declared, though no mainland-China reachability dependency, so no further deduction.
Good layered architecture (SKILL.md trunk + STYLE.md + topic references + task routing), MIT license, version 1.0.0, maintainer and contact metadata. Deducted for missing changelog and explicit update path; references content invisible so progressive disclosure completeness and known-limitation disclosure cannot be confirmed; publisher unverified by the registry.
Output contract is explicit (migrations, code, tests, risk notes, source notes) and matches the stated goal of writing Rails code matching existing conventions, with a plausible marginal value over generic model behavior. Deducted because static review provides no representative output verified as directly usable, and README marketing language substitutes for comparative-benefit evidence.
Positive signals: MIT license, structured version metadata, attributable third-party source (37signals style guide with link), and a declared validation script plus CI workflow. Deducted because the validation script and test suites are not included in the evidence, the CI workflow is only a PR skill-review (not test coverage), and no third-party execution evidence allows independent reproduction of key claims.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →