| name | ux-writing |
| description | Router for UX writing: product voice, microcopy, error/validation messages, settings IA, option labels, feature names, plan names, and button labels. Use when the words appear in the UI. Long-form prose → write-clear-prose; onboarding flow/empty-state strategy → user-onboarding. Read one named reference. Triggers: error copy, rename, label, voice tone, settings copy. |
UX Writing
For a focused question, read the single reference file that matches it — each carries the full distillation for its area (deeper worked examples and sources live in a same-named subdirectory next to it). When a task spans areas (e.g. writing copy for a whole flow needs voice and naming and error states), follow the routing chain below and read each as needed; the one-file rule optimizes Q&A lookups, not multi-surface work.
- Voice, tone & microcopy — defining a product's voice, dialing tone per moment, writing or reviewing general UI microcopy, empty state copy, PACE, localization-ready writing: references/ui-voice-and-tone.md; worked exercises, before→best ladders, tone drills: references/ui-voice-and-tone/voice-and-tone-examples.md; provenance: references/ui-voice-and-tone/sources.md
- Error messages — writing or reviewing any failure-state copy: error toasts, dialogs, form validation, 404/500 pages, retry prompts, payment/auth failures, incident communication: references/error-messages.md
- Settings IA & copy — Linear-style settings pages, subpages, sidebar taxonomy, section groupings, row titles, helper descriptions, dropdown option and action labels: references/linear-settings-copy.md; the underlying Linear settings audit (sidebar IA, page archetypes, row anatomy): references/linear-settings-copy/linear-settings-copy-ia.md
- Naming features & labels — crafting and critiquing names for features, settings, buttons, tabs, modes, plan tiers; descriptive vs clever/branded; candidate generation and elimination: references/naming-features-and-labels.md; full case studies (Apple Cash, Visited Places, Memories, Enhance Dialogue, AutoMix): references/naming-features-and-labels/worked-examples.md
- Sources and gaps — provenance and parked candidate rules: references/sources.md, references/coverage-gaps.md
Internal deference chain (route between references the same way the original skills deferred):
- Voice/tone sets the overall register; when the task is choosing the actual word for a feature, button, value, or plan, route from ui-voice-and-tone to naming-features-and-labels.
- Voice/tone sets where the serious/clinical dial sits in general; any "write/review an error" task routes from ui-voice-and-tone to error-messages, which owns the failure-state ruleset.
- Settings work composes both ways: pick the word with naming-features-and-labels, then place it and write the row with linear-settings-copy; keep its quiet register consistent with ui-voice-and-tone.
When copy becomes agent guidance: if the task is to make agents repeatedly apply a copy decision, name the rule with a stable ID, scope it to a surface, cite the owning reference/source, and include bad/good examples. Keep product judgment in this skill; route packaging, evals, and repo-agent trigger mechanics to creating-skills and working-with-ai.
Boundary splits to respect:
- Long-form prose →
write-clear-prose — essays, articles, emails, sentence-level prose mechanics. This skill owns the words users read in the UI.
- Onboarding flows and empty-state strategy →
user-onboarding — first-run flows, welcome copy, "no data yet" empty states for new users. Failure-state empty screens ("we couldn't load your projects") stay here in error-messages.
