aesthetic-eval

Design review is mostly vibes. This is an attempt to make some of it measurable: point it at a URL and it scores the page across ten UI and UX dimensions against thresholds pulled from published research rather than personal taste.

Browser mode drives a real page through Playwright and runs axe-core alongside the scoring pass, so accessibility and design findings come out of one run. Static mode skips the browser and analyses source, fast enough for a pre-commit hook.

Results emit as SARIF, so GitHub Code Scanning ingests them with no adapter. Baseline diffing is the part I use most: snapshot a passing state, and every later run reports only what regressed against it.