Where does agent-based test fixing actually break down?

Discussions
,
1

I’m building e2epilot — an agent setup that takes a failing E2E suite, probes the app to verify the root cause, writes a patch, and re-runs. Playwright, Cypress, Selenium.

Not here to pitch it. Here to find out where it breaks on real repos.

Real run, no marketing cuts: https://www.youtube.com/watch?v=8vEZOyrjeLA

There is free trial, so… tear it apart.

4

Is this similar to playwright healer agent?

Are you looking at the source code or similar to playwright dynamically running the app.

Are you patching the app or just the tests to make them work?

5

No, E2E Pilot is a standalone agent that works across Playwright, Cypress, WebdriverIO, Selenium (JS/Java), and Seleniumbase — and it does more than healing:

  • Writer: explores your app live in a real browser, then writes new tests from scratch using your existing POMs/fixtures/factories.
  • Fixer: diagnoses and repairs failing tests.
  • Architect: scaffolds a fresh project (config, fixtures, factories, a smoke test) if you don’t have one yet.

Both source code and dynamic runtime. It reads your test files, page objects, and fixtures, and it also drives the app live — capturing DOM snapshots, network, console, and a per-action timeline — so fixes are grounded in what actually happened, not guessed from a stack trace.

Tests only, never the app. E2E Pilot patches test code, page objects, fixtures, and factories. It never touches your application source. If a test is failing because the app is genuinely broken, it gives you a report of the bug