Requirement
A page has one primary focal region, and its visual reading order matches the task's priority order. The eye lands first on the region the user needs to act on first; everything else steps down in size, weight, or position. Two regions of comparable weight with no task reason for the tie is a finding — as is a first-read that lands on secondary content (a decorative summary card outranking the data-entry surface the task requires).
This promotes the "Compose, don't fill" prose (design skill, Phase 2) into a checkable whole-page composition rule.
Rationale
Composition is how a page explains itself before it's read. When the most important thing reads first, the user's eye is doing the work the hierarchy intends; when two regions compete, the reader hesitates, and the page feels assembled rather than composed. A page can pass CMP-5 (one filled button) and SLP-6 (real type-scale steps) and still fail this — the outcome (which region leads) is not guaranteed by either mechanism alone.
How to verify
Judgment. No deterministic sub-check — "which region reads first" is not mechanically scorable without a gaze model.
- Squint test: blur or step back from the 1280 screenshot — which region draws the eye first? Compare to the task's stated priority (Phase 1's primary task).
- Region enumeration: list the page's distinct visual regions (not DOM elements) and
rank them by apparent weight — size, colour saturation, position (top/centre draws first),
whitespace isolation. A finding is two or more regions within the same rank with no task
reason. Align this read with the pattern inventory (
../layout-patterns.md, beside the catalogue) item 1 rather than inventing a separate procedure.
Passes when
- The first-glance read lands on the task's priority region; everything else steps down.
- A page with exactly one region by design (a single-decision confirmation card).
- A deliberate two-panel comparison where the task itself is side-by-side comparison (no single priority region — the "no task reason" clause's exception).
Fails when
- Two or more regions compete for first read with no task reason for the tie.
- The squint-test first-read lands on secondary content.
- The primary action's region is visually subordinate to decoration.
Evaluator guidance
Do not flag: a deliberate multi-region comparison view where the task has no single
priority region; a single-region-by-design page; a reasoned deviation carrying an inline
dx-waive LAY-7 reason="…" (L2).
Deconfliction. CMP-5 is a component rule (one filled button per view); this is the whole-page composition outcome. SLP-6 is one mechanism (type-scale contrast) hierarchy can use; this is the result. LAY-3 says which template shell to use; this says what leads inside it once chosen.