e2e-check

Run E2E tests or perform interactive browser-based verification using Playwright.

Mode Detection

Parse $ARGUMENTS to determine mode:

• Runner mode (default): Run the Playwright test suite via CLI
• Interactive mode: If --interactive or interactive appears in args, use Playwright MCP tools for manual browser walkthrough

---

Runner Mode

1. Detect project

From $ARGUMENTS or current working directory:

• Match project name from cwd (e.g., my-app if cwd contains my-app)
• Check CLAUDE.md in the project root for the configured dev port and test command
• No match: ask which project to target

2. Pre-flight check

Check if the dev server is running on the expected port (read from CLAUDE.md or the project manifest):

Use: bash -c "curl -s -o /dev/null -w '%{http_code}' http://localhost:{port}/ 2>/dev/null" or equivalent.

If the server is NOT running:

• Report: "Dev server not running on port {port}. Start it with your project's dev command (e.g. yarn dev/npm run dev, python manage.py runserver, bundle exec rails s, go run, etc.), then re-run /e2e-check."
• Do NOT auto-start the dev server. Stop here.

3. Run tests

Run the E2E test command from CLAUDE.md or the project manifest (e.g., yarn test:e2e or npm run test:e2e for JS, or pytest tests/e2e for Python).

If a test filter was provided (e.g., auth), resolve it to the spec file path:

• auth → tests/e2e/auth.spec.ts
• Full path → use as-is

4. Parse output

Extract from Playwright output:

• Total, passed, failed, skipped counts
• Duration
• On failure: spec file name, test name, error message

5. Report

E2E Test Results -- {project}
================================
Total: X | Passed: X | Failed: X | Skipped: X
Duration: Xs
Baseline: {expected from CLAUDE.md} tests

[If failures:]
FAILURES:
- {spec-file} > {test-name}: {error summary}

[If screenshots saved:]
Screenshots: test-results/{spec-file}/

6. On failure

• Check test-results/ directory for saved screenshots
• Analyze failure type: timeout? Element not found? Assertion mismatch? Network error?
• Suggest fix based on the failure pattern