E2E Tests: $ARGUMENTS

Creates comprehensive Playwright end-to-end tests in an isolated git worktree.

Arguments

• •Specific feature: login-form, dashboard, settings
• •Glob pattern: user-*, *-modal
• •All features: . or all

Process

1. Create Isolated Worktree

TIMESTAMP=$(date +%Y%m%d-%H%M%S)
WORKTREE_PATH="../e2e-test-worktree-$TIMESTAMP"
git worktree add "$WORKTREE_PATH" -b "e2e-test-$TIMESTAMP"

2. Setup Playwright

If not installed:

npm install -D @playwright/test
npx playwright install --with-deps

Create playwright.config.ts if missing with standard configuration.

Add npm scripts if missing:

{
  "test:e2e": "playwright test",
  "test:e2e:ui": "playwright test --ui",
  "test:e2e:debug": "playwright test --debug"
}

3. Identify Features

• •For specific feature: Create single test file
• •For glob: Search components/app directories, create tests for matches
• •For all: List all components/routes, use Task tool with parallel subagents if >3 features

4. Generate Tests

Create tests/e2e/[feature-name].spec.ts with:

• •Happy path scenarios
• •Edge cases and error states
• •Input validation
• •Accessibility checks

Selector strategy: Use data-testid attributes. Add them to components if missing.

5. Run and Fix

npm run test:e2e

Application fixes (MUST be general, not test-specific):

• •Add missing data-testid attributes
• •Fix broken functionality
• •Add proper loading states

Test fixes:

• •Update selectors
• •Fix assertions
• •Add proper waits

6. Validate Full Suite

npm run build
npm run lint
npm test
npm run test:e2e

7. Report and Prompt

On success: Prompt to merge or keep for review On failure: Categorize errors, offer debug or cleanup options

Test Standards

• •Use Page Object Model for complex features
• •Test coverage: happy path, edge cases, errors, accessibility
• •Prefer data-testid over CSS selectors
• •Use toBeVisible() not arbitrary timeouts
• •Tests must be deterministic and independent

Parallel Execution

When >3 features to test:

• •Split into groups of 1-2 features per agent
• •Launch Task tool with general-purpose subagents
• •Collect results and continue with validation