nirarad/playwright-ai-qa-agent — reverse-engineered prompt

Build me a small TypeScript project that shows an AI powered QA workflow end to end. I want a simple task manager demo app I can run locally or deploy, with a few built in break modes so I can simulate a selector change, a real logic bug, slow timing, and a broken login.

Add Playwright tests for the main user flows, and make sure test runs save JSON results, an HTML report, screenshots, traces, and videos when things fail. Then add a local failure agent that reads the Playwright output and uses a configurable LLM, plus a mock mode, to classify each failure as a broken locator, a real bug, flaky, or an environment issue.

If confidence is high, have it open GitHub issues for automation bugs and real product bugs, and if possible include an optional healing PR flow for broken locators. Flaky and environment problems should just be logged. Please wire the whole thing into GitHub Actions so the agent still runs after test failures, artifacts are preserved, and the workflow still ends red when tests fail. Look up current docs online if you need to.