What problem does it solve? Comparing a Flutter implementation against a design mockup by pixel identity always fails because Chrome and Skia rasterise text, shadows, and gradients differently, so teams either disable visual checks or argue endlessly over what counts as a defect. This Skill defines a measurable parity standard: everything a human decided (tokens, layout order, copy, RTL mirroring) must match exactly, everything a rasteriser decided is compared within stated tolerances (±2 logical px, ΔE00 ≤ 2), and renderer artifacts are declared legitimate differences. ## Core Features & Use Cases - Three-tier parity standard: exact-tier assertions on token values, element order, state signals, copy, and RTL geometry; tolerance-tier measurements via getRect and pixel sampling; a differs tier that is never a finding. - Reproducible capture pipeline: headless-Chrome regeneration of the four 2600x4760 reference sheets, a crop-rect dump via --dump-dom, zero-dependency cropping, and Flutter captures at 390x844 logical / DPR 2 with loadAppFonts, pinned clocks, and seeded fixtures. - Full comparison matrix: 24 paired captures across six screens x {light, dark} x {en/LTR, fa/RTL}, plus no-reference passes for German longest-string overflow, Sorani script rendering, and 200% text scale. - Use Case: Before merging a UI epic, capture the four parity cells per screen, build the side-by-side sheet with a difference column, paste it into the PR description, and back it with widget-test assertions on resolved token values and measured rects. ## Quick Start Use the daybreak-visual-parity skill to capture the Today screen at 390x844 DPR 2 in light and dark for en and fa, crop the matching reference frames, and produce the side-by-side parity sheet for my PR.
