Reference
Design system
Every ratio was computed with tools/a11y/contrast.py, not eyeballed. This page is
noindex: it documents the template for whoever builds with it.
Colour tokens
--color-bg
#F5F5F7 — Cool pale ground
--color-bg-alt
#E4E4EA — Banding
--color-text
#15161C — 16.58:1 on ground · AAA
--color-text-muted
#54555F — 6.79:1 on ground · AA
--color-band
#15161C — Inverted band
--color-on-band
#FFFFFF — 18.05:1 on band · AAA
--color-on-band-muted
#AEAFB8 — 8.27:1 on band · AAA
--color-accent
#2B3480 — Indigo, hue 234° — 11.02:1 with white on top
--color-accent-text
#262E72 — 11.24:1 on ground · AAA
--color-accent-on-band
#8E97E8 — 6.65:1 on band · AA
--color-required
#262E72 — Decoration; the asterisk is announced in words
The hue was chosen by elimination
All 88 accents shipped before this were bucketed by hue in 15° steps. 225–239° was the only empty bucket in the library, so this one lands there. The first candidate was a petrol teal and it was discarded for sitting 4° from day 84 at identical lightness — which is the "same template, different colour" failure this process exists to catch.
Readings use real <meter>
min, low, high and optimum are that element's own attributes, so a value and its bounds are announced without a line of ARIA. Every panel also states its verdict as a word, so the meter is never the only route to "do not polish this".
Correction available Do not polish
Type
Gantari — display
Sometype Mono — 31 µm · 140 µm · 45 µm
The readings are the argument, so they are tabular and monospaced and the step between a factory panel and a repainted one is visible at a glance. Neither face appears in days 1–88.