flakeproof
When CI goes red, flakeproof tells you whether the app broke or just the test.
The problem
An end-to-end suite can fail you in two ways. Blind tests stay green while something real breaks, fragile tests go red because a class name changed. Both get called flaky, and every red build turns into guesswork.
What I built
A tool that mutates the page under test on purpose and watches how the tests react. Meaningful changes, like a changed button text, must turn a good test red. Cosmetic ones, like a renamed class or an extra wrapper, must not, and neither should an element that appears 800 ms later. When CI fails, flakeproof compares the failing selector against a baseline and gives a verdict.
What came out
Four verdicts instead of a shrug, real change, fragile, nondeterministic or unclear. For fragile selectors it proves a better one instead of guessing it, and when the evidence is mixed it says so.
Built with
Node.js, Playwright, works with Playwright, Robot Framework, Cypress, Selenium and Puppeteer