What problem does it solve? Frontend fixes are often declared done based on reading code, but browser behavior routinely contradicts what the code suggests — CSS rules get silently overridden, hidden images still download, and test servers keep serving stale code. This Skill replaces inference with measurement: it drives a real browser, verifies the server serves what is on disk, and proves the test itself can fail before trusting a pass. ## Core Features & Use Cases - Verified server restart: The restart_and_verify.sh script restarts a local test server and fails loudly if the served content does not match the edited file on disk, catching stale template caches and silent port conflicts. - Real input measurement: Uses CDP Input.dispatchTouchEvent for genuine touch gestures, network request counting, and sanity controls (scrollability, element presence, actual touch targets) before any claim. - Control runs and mutations: Re-runs the same test against pre-fix code and applies mutations to prove each assertion is capable of failing, turning a passing test into actual evidence. - Use Case: After fixing a CSS bug in a Flask app, restart the server with verification, measure the behavior in Chromium via Playwright, confirm the same test fails on the old code, and only then report the fix as working. ## Quick Start Ask the AI to verify your frontend fix in a real browser using the browser-measurement discipline before reporting that it works.
