界面风格方案对话选型

SKILL开发

原名 oh-my-ui-style

开发 未标注权限 许可证不明确
安装这个技能

先确认做整站管理平台还是单页,再在对话中给出可点击的界面方案,先定整体框架再定配色,让用户不离开聊天就能选定视觉方向。

一句话选型结论

按资料字段整理

界面风格方案对话选型是开发类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类开发 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_41854512/oh-my-ui-style
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

开发
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_41854512仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验2026-11-06上游最后核验时间 · 数据更新 2026-11-06
正文体量7.3 KBSKILL.md 原文大小,越大说明说明书写得越细

能做什么

0 项能力
上游给出的能力概览与适用场景
选定后台界面风格对话内预览配色确定页面框架结构

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写代码与改 bug

按需求补代码、定位并修掉报错。

读仓库与生成改动

让模型读懂项目结构,再给出可落地的改动。

跑测试与排查故障

按现象排查问题,补测试用例并复跑。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在开发场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
界面风格方案对话选型当前 未标注权限 未标注 未标注 未标注 有
技能审查员 浏览器控制 ClawHub 市场 877.9 万 OpenClaw 有
Vercel React 最佳实践 浏览器控制 GitHub 仓库 849.1 万 Codex、Claude、Cursor 有
GitHub 代码协作 联网访问 ClawHub 市场 763.3 万 OpenClaw 有
网页设计指南 联网访问 GitHub 仓库 668.6 万 Codex、Claude、Cursor 有
本体论 只读 ClawHub 市场 663.7 万 OpenClaw 有

SKILL.md 全文

7.3 KB
模型实际读到的说明书原文,来自技能仓库

Oh My UI Style

Turn "make it look good" into a short sequence of clickable decisions inside the conversation, then hand the chosen tokens to implementation.

Scope

Use for structure, style, palette, and layout-direction decisions on admin consoles, dashboards, internal tools, and single pages. Skip when the project already has a design system to follow, or when the user asked for the finished page - implement that directly.

Step 0 - Ask the mode first

Ask one short question before generating anything:

  • 整体管理平台 - login page, navigation shell, header, theme tokens, and the content area.
  • 单个页面 - one page's own structure and styling inside an existing shell.

Rules:

  • Skip the question when the request already states the mode ("整个后台", "这个列表页", "just this page").
  • Ask at most one clarifying question. Infer modules from the project; when the user names them, use their names verbatim.
  • Announce the mode in one line before rendering.

Design source

Decide where the design direction comes from before writing any tokens.

  1. Check the skills already loaded in this session for a design-direction skill: frontend-design, ui-ux-pro-max, design-*, brand-*, or any description covering visual design, design systems, or UI style.
  2. If none is loaded, scan the frontmatter descriptions of ~/.codex/skills/*/SKILL.md, ~/.agents/skills/*/SKILL.md, and ~/.config/magpie/library/skills/*/SKILL.md for the same keywords.
  3. When a fitting skill is found, read it and let it own palette, type roles, layout concept, and signature element. This skill then owns only three things: variants must differ structurally, the picker contract, and the selection handoff.
  4. When a design skill exists but does not fit this product - another brand's guidelines, a different medium - say why it was skipped and continue.
  5. When nothing fits, use the built-in fallback below.
  6. Name the design source in the final reply ("direction from frontend-design" or "built-in fallback"). Never present freehand model choices as if they were a project standard.
  7. If a known design skill such as ui-ux-pro-max is missing and would clearly improve the result, mention that skill-installer can add it.

Built-in fallback

Use only when Design source found no fitting skill.

  • Lock the brief in one pass: page name, the 3-4 core modules, framework or design-system constraints, light or dark or both. Infer defaults from the project where possible; ask at most one round of questions.
  • Write the token plan first: 4-6 named hex values (surface, raised surface, border, text, primary, one semantic color), 2 type roles (display + body, plus a mono utility role for data-heavy screens), a one-sentence layout concept, and one signature element.
  • Re-read the plan against the brief. If a proposal would fit any similar brief unchanged, revise that proposal before rendering.

Variants

  • Whole platform defaults to 5 shell proposals; a single page defaults to 4.
  • Both counts are starting points, not limits. When the user asks for more, or the information architecture is deeper than the default set can express, extend to 6-8 and invent archetypes grounded in the project's real menu structure.
  • Each variant must differ in layout and interaction, not just hue. Real archetypes to draw from: equal-weight quadrant grid, sequential pipeline, master-detail split, dense telemetry HUD.
  • Name variants with descriptive words ("Bento 矩阵", "Pipeline 流水线"); never prefix them with numbers.
  • Fill each mini interface with the project's real nouns - module names, field names, severity levels - not placeholder text.

Mode A - Whole admin platform

Two steps, both rendered in the conversation.

Step 1, shell structure:

  • Cover the shell explicitly: login, navigation, header/topbar, content area, theme tokens.
  • Default seeds: icon rail, standard sidebar, top navigation, grouped sidebar, hybrid navigation. Treat this list as examples, not a closed set - add or replace shells to match the real menu depth.
  • Every card shows: a short id, the name, a wireframe preview drawn in the dark ops example palette (real hex values, not grey placeholders), one line on which platform shape it fits, and the card's example tokens (surface / panel / primary / alert).
  • Single select, with a persistent "current selection" bar and one confirm action.

Step 2, palette and type:

  • Take the selected shell and render 3-4 palettes on that same shell. Do not change the structure.
  • Each proposal shows the swatch strip, the type pairing, and the same shell preview in that palette.
  • Single select again; then hand off the combined tokens.

Mode B - Single page

  • Default 4 proposals; more when asked.
  • Keep the surrounding shell constant and vary the page itself: list + detail, table-first, card wall, split editor, dashboard.
  • Each proposal includes its own dark palette preview, swatch strip, and type pairing.

Rendering

Read the visualize skill and follow its fragment contract exactly before authoring.

  • If the visualize skill is available, always render in the conversation. Never default to writing an HTML file the user must download or open separately.
  • Write files only when the user explicitly asks for a file, or when no in-conversation surface exists; then fall back to imagegen mockups, and only then to open_in_codex.
  • When a design skill's aesthetic rules conflict with the token guidance here, the design skill wins. The fragment contract, picker structure, and selection handoff always follow this skill and the visualize contract.
  • Give the fragment root a unique id and scope every selector to it.
  • Selection is single-select: clicking a card marks it selected, updates the bottom bar, and saves window.openai.setWidgetState({ selectedScheme }).
  • The confirm button calls window.openai.sendFollowUpMessage with the full token string. Guard both host APIs.
  • Bind Tweak where it adds something the picker does not already choose - for example the example accent inside the shell picker. Skip it when the control would duplicate the choice itself.
  • Reflow to one column below 640px instead of shrinking the wireframes.
  • Start from assets/platform-shell-picker.html for Mode A step 1 and assets/style-picker-template.html for step 2 or Mode B; replace placeholders instead of restructuring the shell.

After the pick

  • Restate the chosen tokens, write them as one CSS variable block in the project, and require later pages to consume only those variables.

Quality bar

  • Body text keeps at least 4.5:1 contrast on its own surface; state colors always pair with text or shape.
  • Motion only where it encodes state, and honor prefers-reduced-motion.
  • Keep one design source for all variants; never mix a design skill's direction with freehand fallback choices in the same set.

常见问题

5 问
答案由本站依据条目字段整理
界面风格方案对话选型怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_41854512/oh-my-ui-style。安装方式为未标注。装完刷新技能列表或重开客户端。

界面风格方案对话选型支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

界面风格方案对话选型权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

界面风格方案对话选型收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

界面风格方案对话选型用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 2026-11-06。