Requirement
Descriptive copy leads with its purpose — what it does for the user and when to reach for it — before the mechanism. State the why first; the what (the tool, token, or library) follows, or lives in the body. A reader should know why a thing exists, and when they would need it, from its first line.
This binds hardest on descriptive prose: page titles, descriptions, section intros, empty states, and feature blurbs — the surfaces where it is easy to state the what and skip the why. Instructional copy and running body prose are governed by CNT-3 (voice mechanics), not this control.
Rationale
People scan descriptive copy to decide whether a thing is for them. A description that opens with the mechanism ("shadcn/ui default token scales, unmodified") makes the reader do the translation from implementation to value; purpose-first copy does that work for them.
Split out of CNT-3 on 2026-07-03, during the writing-guide port: the purpose clause is pure judgment while CNT-3's other clauses (person, voice, sentence length) are mechanical, and the bundle kept straddling guideline sections — the mechanics under "Concise"/"Conversational", this clause under "Clarity". One ID for both made findings ambiguous and meant a waiver on sentence length could silently cover a purpose failure. One control, one verifiable failure mode.
Passes when
- A description or intro opens with its purpose: "Consistent spacing is what makes a screen feel calm instead of busy. One shared scale." — the mechanism (the token scale) comes second.
- An empty state opens with what the user can do next, not with the name of the feature that is empty.
- A feature blurb's first line answers "why would I use this?".
Fails when
- A title, description, or intro opens with the mechanism (the tool, token, or library) instead of what it does for the user: "shadcn/ui default token scales, unmodified" as a section's first line.
- A page description that only names the page's parts, with no line about who it serves or when to reach for it.
How to verify
Judgment only — there is no deterministic half. The evaluator reads the first line of each title, description, section intro, empty state, and feature blurb in scope and asks: does this state the purpose or role before the mechanism?
Evaluator guidance
Quote the offending first line in every finding, and give the purpose-first rewrite.
Flag:
- Mechanism-first descriptive copy: a title, description, or section intro whose first clause names the tool, token, library, or data structure instead of the purpose or role it serves.
- A description that is only a scope list ("Labels, buttons, helper text") with no purpose line anywhere.
Do not flag:
- Mechanism named first when the mechanism is the point: a spec table, an API reference, or a row whose job is to state a value ("Radius: 8px"). This control governs prose that introduces or explains, not reference data.
- Body prose that explains the mechanism after the purpose has been stated — the control binds on the first line, not the whole passage.
- Settled product names opening a line — an identity is not a mechanism.
Waiver
rationale (L2) — inline dx-waive CNT-7 reason="..." at the deviation site.