Running body text is held to a comfortable measure — roughly 45-75 characters per line (target 40-60), via a max-width in ch, not full-viewport paragraphs

Requirement

Hold running body text to a comfortable measure — the number of characters per line. Target 40–60 characters, and treat roughly 45–75 as the acceptable band. Cap the width of prose blocks with a max-width expressed in ch (Tailwind's max-w-prose, or an explicit max-w-[65ch]), rather than letting paragraphs stretch the full viewport on a wide screen.

This is running prose only — instructional text, help copy, descriptions, empty states, long-form guideline and content pages. Headings, table cells, data grids, and short labels are out of scope; they have their own layout logic and are not held to a measure.

Rationale

When a line of prose runs much past ~75 characters the eye has to travel too far to find the start of the next line, and long blocks become a wall of text; much under ~40 and the reader hyphenates and re-fixates too often. Both cost reading speed and comprehension. A single max-width in ch scales the cap to the actual font, so the measure stays comfortable across breakpoints without hard-coding pixel widths — Utility by Default, and easy on the reader. It is a strong default with reasonable exceptions (a two-column layout, a deliberately narrow aside), so it is L2.

Passes when

• A paragraph of body copy sits inside a max-w-prose (or max-w-[65ch]) container, so it wraps at a comfortable measure regardless of viewport width.
• A long-form content or guideline page constrains its prose column rather than running text edge to edge on a wide monitor.
• A table, data grid, or heading spans its natural width — no measure cap needed, and none expected.

Fails when

• A paragraph of running prose spans the full viewport width with no max-width cap, so on a wide screen each line runs well past ~75 characters.
• Body copy set in a container wide enough that its lines routinely exceed ~75ch.

How to verify

Hybrid. The checks/type-scan.py measure rule, built in #161, narrows the surface: it judges ch caps already written against the 75ch ceiling, and treats sub-45ch values as headings and labels rather than prose. It shares one rule body with LAY-4, which caps the same thing at 80ch, so a measure past 80ch reports under both ids. Where no cap is written, or the cap is not in ch, the rule stays quiet (or prints a NOTE) and the measure falls to the rendered output: verify at a wide viewport and label it "verified manually". The evaluator judges the remainder: which blocks are genuinely running prose (and so need a measure cap) versus headings, tables, and data grids (which do not).

Evaluator guidance

Flag: a block of running prose whose lines run the full width of a wide viewport with no measure cap — check the wide-viewport screenshot for paragraphs whose lines are clearly longer than a comfortable measure.

Do not flag: headings, table cells, data grids, code blocks, or short labels; a deliberately full-width layout element that is not running prose. This is L2 — a reasoned deviation (e.g. a two-column reading layout, or a narrow aside below the target) takes an inline dx-waive TYP-6 reason="…"; "forgot to cap it" is not a reason.