用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/johnlindquist/uxspec --skill clarify命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | clarify |
| description | Improve every piece of text in a UXSpec — descriptions, labels, error messages, empty states, and loading copy |
Improve every piece of text in a UXSpec — $description fields, button labels, error messages, empty state copy, loading text. Words are design.
When reviewing a spec's text content. When $description fields are vague. When error states say "An error occurred." When empty states are blank.
These are the most-read part of a spec — they appear in PR reviews, agent prompts, and design handoffs. Write them like you're briefing someone who will build the UI without seeing a mockup.
Include: layout direction, spacing character, element hierarchy, color intent, emphasis, interactive affordances.
Explain what happened. Suggest a fix. Never blame.
These are opportunities, not dead ends. Describe what will appear. Explain the value. Offer a clear action.
Describe the action specifically. Active voice. Match the user's mental model.
Set expectations. Explain what's happening.
Confirm what happened. Guide to next steps.
State the specific action. Explain consequences.
Match the brand personality established in the gather phase:
Be consistent within a spec. If errors are friendly, success messages should be too.
Every string in a UXSpec is an interface with a human. Vague copy creates vague implementations. Specific, intentional text — even in $description fields that no end user sees — produces specific, intentional UIs.