Act as a test engineer. Add end-to-end browser tests to my {{tech_stack}} project for the flows that matter most: {{features}}. Use the end-to-end tool already in the repo, or Playwright if there is none. Read the code first and show me the list of scenarios before writing tests. For each flow, cover the happy path plus the most likely failure, such as a wrong password, a declined test card or a missing required field. Rules: - Find elements by role and label, the way people do, not by CSS classes; if something can't be found that way, flag it as an accessibility issue - No fixed waits: wait for what the user would see - Each test creates its own data and cleans up, so tests can run in any order - Seeded test users for each role, with credentials in environment variables - Payments in test mode and emails caught by a test inbox or mock; never touch live services - Run each flow at a phone size and a desktop size - An automated accessibility scan on each page the tests visit - A trace and screenshot saved when a test fails Add a script to run them locally and a CI job that runs them on every pull request, aiming for under five minutes. When you're done, list the scenarios covered, what isn't covered yet, and any flaky steps you noticed.