The three agents that npx playwright init-agents installs, hosted in your own workspace against aut.dionny.dev. Every run gets a cloud browser you can watch on the left.
- Plan: the planner explores the site and writes
specs/checkout.plan.md.
- Generate: the generator turns plan items into tests under
tests/.
- Run: a failing total is the app's bug, not the test's.
- Break: rename a locator in a passing test, like a UI change would.
- Heal: the healer repairs it while a guard watches the expected values.
- Review: an open-source review skill critiques the test.
Stages can be revisited and redone at any time. Running behind? Use the sample plan or sample tests.