Natural product-interface microcopy for labels, CTAs, states, help text, onboarding, settings, trust/privacy disclosures, landing-page interface copy, and locale-specific UI wording under frozen product meaning. Use for browser extension, web, desktop/WebView, mobile, and app surfaces after product semantics and UI role are known.
Resources
2Install
npx skillscat add yuukias/ai-skills-collection/product-ui-copy Install via the SkillsCat registry.
Product UI Copy
Use this skill for short user-facing product-interface copy: labels, buttons,
calls to action, empty/loading/error/success states, settings text, onboarding,
help text, trust/privacy disclosure, permission prompts, landing-page interface
blocks, and other copy that appears inside a product surface.
This is not the long-form Chinese prose route. Reports, README files,
technical notes, scientific explanations, and ordinary Chinese document
polishing stay with chinese-prose or scientific-rewrite as appropriate.
Ownership
Product/domain/legal authority owns product truth:
- availability and eligibility;
- current product state;
- user consequence and next action;
- payment, subscription, pricing, cancellation, deletion, retention, consent,
privacy, safety, and trust facts; - exact product, brand, feature, technical, and policy identities.
Frontend Design owns content architecture:
- whether text belongs on the surface;
- placement, amount, hierarchy, neighboring copy, duplicate payload, and
progressive disclosure; - final rendered page or screen rhythm.
Product UI Copy owns wording under frozen meaning:
- concise natural wording;
- locale/register;
- first-pass linguistic page rhythm;
- preserving or escalating protected meaning.
If product truth or legal/trust meaning is missing, do not invent it. Return a
clear escalation instead of a polished guess.
Required Handoff
When receiving work from Frontend Design, expect this lightweight handoff:
SURFACE:
UI_ROLE:
PRODUCT_STATE:
USER_JOB:
USER_CONSEQUENCE_OR_NEXT_ACTION:
NEIGHBORING_VISIBLE_COPY:
LOCALE:
PROTECTED_MEANING:
DISCLOSURE_LEVEL:
LENGTH_OR_VIEWPORT_CONSTRAINT:
DESIGN_AUTHORITY: optional
TERMINOLOGY_OR_BRAND_TOKENS: optionalDo not turn this handoff into a schema, database, ledger, or state machine.
Decision Classes
Use these labels as reasoning structure when they help keep ownership clear:
KEEP: the current copy is already clear and should not be rewritten just to
show activity.WORDING/NATURALNESS: wording can be improved without changing meaning.LOCALE/REGISTER: the same protected meaning needs locale-specific wording.CONTENT ARCHITECTURE: Frontend should adjust placement, amount, hierarchy,
disclosure, or neighboring text before wording is final.PRODUCT SEMANTICS: product truth, state, eligibility, consequence, or
reversibility is missing or inconsistent.LEGAL/TRUST/SAFETY: consent, privacy, retention, deletion, safety, pricing,
visibility to others/systems, or trust claims need product/domain/legal
authority.
Locale
zh-Hans and zh-Hant-HK are independent realizations from one protected
meaning. Character conversion alone is not localization.
For zh-Hant-HK, prefer Hong Kong user-facing wording and register where it is
meaningfully different. Preserve exact product names, feature names, technical
tokens, and policy terms that must match the product.
Workflow
- Confirm the handoff or prompt includes enough product truth to write safely.
- Classify each requested line as
KEEP, wording, locale, content
architecture, product semantics, or legal/trust/safety. - Mark protected meaning before rewriting. Do not change user consequence,
eligibility, privacy/safety/trust meaning, pricing, reversibility, or exact
product identity. - Produce the smallest set of copy changes that improves the interface.
- For locales, realize the same protected meaning independently for the
target locale. - Return unresolved product/legal/trust questions explicitly.
- Ask Frontend Design to validate final rendered rhythm when layout, viewport,
density, or neighboring copy can change meaning or readability.
Output Standard
- Short copy sounds like a product surface, not a report, audit log, or
internal implementation note. - Buttons use action verbs that match the real consequence.
- Help and trust text says what the product actually does and does not hide
missing facts. - Empty/error/loading/success states tell the user what happened and what they
can do next. - Already-good copy can remain unchanged.
- The copy does not over-explain just because there is available space.
- The result makes clear which owner must act next when wording is blocked by
missing product truth or content architecture.
Red Flags
- A UI request routes to long-form
chinese-prosejust because the text is
Chinese. - Wording hides unresolved privacy, legal, safety, payment, deletion, or
visibility semantics. - The copy changes an action consequence from reversible to final, public to
private, optional to required, or diagnostic to user-facing. - A locale version is only character conversion.
- All lines are rewritten even when some should be
KEEP. - Frontend layout or disclosure decisions are presented as Writing decisions.
- A backend/API/docs-only request activates this skill without a product
interface surface.