rails-cto-qa

The quality gate. The most important skill in the plugin — every Ruby file must pass its three gates before it's considered done.

What it does

Runs your Ruby changes through three gates, in order:

1. Linting

• Static analysis via rails-cto-static-analysis — Reek, Flog, and Flay catch code smells, excessive complexity, and duplication before RuboCop runs
• RuboCop with aggressive autocorrect (bundle exec rubocop -A) for style and consistency
• No rubocop:disable escape hatches — fix the code, not the rules

2. Tests

• Runs the tests for the changed files in parallel with coverage enabled
• Requires subject to be declared exactly once per test class
• Reads coverage/coverage.json (generated by the SimpleCov JSON formatter) to evaluate per-file coverage
• Thresholds: ≥ 90% good, 70–89% acceptable, < 70% insufficient — the skill adds tests until coverage is high enough

3. Code review

• DRY check — three similar lines is a signal to extract
• Dead code removal
• Edge case and error handling check
• Performance review — N+1 queries, missing indexes, heavy work in requests
• Documentation check — every method has a comment explaining why
• Debug statement cleanup — no stray puts, pp, or binding.pry
• Clarity check — explicit over clever

Nothing is considered done until all three gates pass.

When it triggers

• After creating or modifying any .rb file — model, controller, command, service, job, test, anything
• When you mention rubocop, lint, style, autocorrect, code quality, code review, QA, or review my code

This skill is mandatory — the CLAUDE.md block installed by rails-cto init enforces it. You cannot ship Ruby changes without it running.

Example

You: Add a cache method to the User model.

Agent: [writes the cache method in app/models/user.rb]
       [invokes /rails-cto-qa]
         → runs /rails-cto-static-analysis (Reek/Flog/Flay)
         → runs bundle exec rubocop -A app/models/user.rb
         → runs the user model tests
         → reads coverage/coverage.json → 96% coverage, good
         → reviews the diff for DRY, performance, error handling
         → flags a missing `# Why:` comment on the new method
         → adds the comment
         → re-runs the gate
         → reports: ALL GATES PASS

View source

skills/rails-cto-qa/SKILL.md on GitHub →