name: rails-system-test-analyzer
description: Analyze Rails system tests to identify which ones could be converted to faster controller tests. Use when optimizing test suites, reviewing system tests, or when asked about test performance.
argument-hint: "[test/system/file_test.rb]"
disable-model-invocation: true
allowed-tools:
- Bash
- Glob
- Grep
- Read

Rails System Test Analyzer

Analyze Rails system tests to determine if they truly require browser-based testing or could be converted to faster controller/integration tests.

Core Principle

System tests should ONLY be used when testing functionality that absolutely requires a real browser:
- JavaScript interactions (Stimulus, Turbo, custom JS)
- Complex client-side state
- Browser-specific features (file uploads with preview, drag-and-drop, etc.)

If a test only verifies server-rendered HTML, form submissions, and redirects, it should be a controller test.

Analysis Process

Step 1: Identify System Tests

If $ARGUMENTS is provided, analyze those specific files. Otherwise, find all system tests:

find test/system -name "*_test.rb" -type f 2>/dev/null || find spec/system -name "*_spec.rb" -type f 2>/dev/null

Step 2: For Each Test, Trace the Code Path

For each test file, analyze each test case:

1. Extract the test actions: Look for visit, click_on, click_link, click_button, fill_in, select, check, uncheck, choose, attach_file

2. Identify routes/controllers: Map visit some_path to controller actions using config/routes.rb

3. Find views rendered: From the controller action, identify:

4. Main view template (app/views/controller/action.html.erb)
5. Partials rendered (render partial:, render @collection)
6. Layouts used
7. Helpers referenced (app/helpers/)

8. Stimulus controllers referenced via data-controller attributes (check app/javascript/controllers/)

9. Check for JavaScript dependencies in each view/partial/helper:

Step 3: JavaScript Detection Checklist

Scan views, partials, and layouts for:

Stimulus/Hotwire (REQUIRES system test):
- data-controller=
- data-action=
- data-*-target=
- <turbo-frame
- <turbo-stream
- data-turbo-method=
- data-turbo-confirm=
- data-turbo-frame=

Traditional JavaScript (REQUIRES system test):
- <script> tags with logic (not just analytics)
- onclick=, onchange=, onsubmit=, onfocus=, onblur=
- javascript: links
- remote: true in form helpers (legacy Rails UJS)

Capybara JS assertions (suggests JS dependency):
- have_css with visible: option
- have_selector with wait: option
- accept_alert, accept_confirm, dismiss_confirm
- execute_script, evaluate_script
- within_window, switch_to_window

Step 4: Classify Each Test

For each test, determine:

KEEP AS SYSTEM TEST if:
- Any JavaScript detected in code path
- Test uses JS-specific Capybara matchers
- Test verifies dynamic content updates
- Test checks client-side validation
- Test involves file uploads with preview
- Test requires specific browser behavior

CONVERT TO CONTROLLER TEST if:
- No JavaScript in code path
- Only tests server-rendered content
- Only verifies redirects and flash messages
- Only checks presence of static elements
- Form submissions without JS enhancement

Step 5: Generate Report

For each test file analyzed, output:

## [test_file_name.rb]

### Test: "test description"
- **Code path**: UsersController#create -> views/users/new.html.erb
- **JavaScript detected**: None / [list findings]
- **Recommendation**: CONVERT / KEEP
- **Confidence**: High / Medium / Low
- **Reason**: [brief explanation]

[If CONVERT, provide controller test skeleton]

Controller Test Conversion Template

When recommending conversion, provide a skeleton:

# test/controllers/[controller]_test.rb

class [Controller]Test < ActionDispatch::IntegrationTest
  test "[original test name]" do
    # Setup
    user = users(:one)

    # Exercise
    get new_user_path
    assert_response :success

    post users_path, params: { user: { name: "Test" } }

    # Verify
    assert_redirected_to user_path(User.last)
    follow_redirect!
    assert_select "h1", "Test"
  end
end

Summary Statistics

At the end, provide:

## Summary

- Total system tests analyzed: X
- Tests requiring system test (JS detected): Y
- Tests convertible to controller tests: Z
- Potential time savings: ~Z * 2-5 seconds per test run

### Quick Wins (High confidence conversions)
1. test_file.rb - test_name
2. ...

### Needs Review (Medium confidence)
1. ...

Important Notes

• Always read the actual test code, views, and controllers - don't guess
• Check both direct JavaScript and inherited from layouts
• Consider that some Turbo features work without JS (progressive enhancement)
• When in doubt, recommend keeping as system test
• Provide actionable, copy-pasteable controller test code