Measure the contrast of every stylesheet this project ships, and darken the one that reads under it

The task inbox writes the assignee cell of an unclaimed task in a muted gray, #767f88, set by .orchestra-inbox .orchestra-assignee--none in modules/orchestra_inbox/css/inbox.css. Against the row it sits on that is 4.07:1, and against the hovered row (#f8fafc, set a few rules above it in the same file) it is 3.89:1. WCAG 2.1 AA success criterion 1.4.3 asks 4.5:1 of text at that size.

So the one cell on the page that says nobody has picked a task up is the cell some readers cannot read, on a listing built to be used inside public bodies.

Why nothing said so

The project already computes this rather than judging it by eye. PresentationAssetsTest::testEveryColoredTextMeetsTheContrastMinimum reads every color pair out of modules/orchestra_presentation/css/cards.css, computes the WCAG relative luminance, and fails under 4.5:1. It also carries a completeness guard, so a rule added to that file cannot ship unmeasured.

It reads that one file. Five stylesheets ship. The rule is stated at one door of five, and the inbox is the door the failure was behind. Every other shipped pair clears AA when computed: the card and status palettes from 4.69:1 to 17.40:1, the instance-detail timestamp at 5.41:1, and the lifecycle dots at 3.40:1 or better, which is what success criterion 1.4.11 asks of a non-text shape.

Proposed resolution

• Darken the muted assignee to #646d76, which keeps its blue-gray cast and reads 5.26:1 on the row and 5.03:1 on the hovered row.
• Lift the stylesheet parsing, the relative luminance and the contrast ratio out of PresentationAssetsTest into a trait both tests use, so one implementation answers the question.
• Give orchestra_inbox a guard of its own over inbox.css, with the same completeness check: a rule that decides how text reads and is named by no case fails the test.

User interface changes

The assignee cell of an unclaimed task in the inbox reads in a darker gray.

AI-Generated: Yes (Claude Code was used to help draft this issue summary and to write the code and tests on the merge request. I reviewed and ran the work myself before posting it.)