Token导航 LogoToken导航

Agent Skill · 研究检索

design-audit(设计审核)

用于辅助界面设计、视觉规范、排版、配色、布局和交互体验优化。它适合让 Agent 根据产品场景整理页面结构、生成 UI 方案、检查视觉一致性或改进组件层级。使用时需要结合现有品牌、设计系统和用户任务,不应只堆装饰元素;涉及真实页面改动时,应通过截图或浏览器预览检查文本溢出、对齐和响应式表现。

design-audit 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 18,846。

总安装 5,880周安装 245GitHub Stars 196下载 1,960核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:design-audit(设计审核)
来源仓库:https://github.com/bencium/bencium-claude-code-design-skill
仓库路径:skills/design-audit
安装命令:
npx skills add https://github.com/bencium/bencium-claude-code-design-skill --skill design-audit
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/bencium/bencium-claude-code-design-skill --skill design-audit

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

design-audit 作为 UI/UX 架构师角色,系统性审查界面设计的必要性与有效性。适用于 Codex、Claude、Cursor、Gemini CLI,聚焦于减少冗余元素,提升操作直觉性。依据设计系统、前端规范与应用流程文档进行判断,不介入功能实现仅关注体验质量。

执行前需提供完整的设计上下文文件,否则难以做出准确可用性评估与改进建议。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/design-audit

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/bencium/bencium-claude-code-design-skill --ski…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

design-audit 怎么安装?

在终端执行 npx skills add https://github.com/bencium/bencium-claude-code-design-skill --skill design-…。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

design-audit 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

design-audit 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

design-audit 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

design-audit 用的人多吗?

公开数据是累计安装 5,880 次,近一周安装 245 次,下载 1,960 次,GitHub 196 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

design-audit 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
design-audit

设计审核 · 本页

5,880196需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Design Audit Skill

You are a UI/UX architect. You do not write features or touch functionality. You make apps feel inevitable — like no other design was ever possible. If a user needs to think about how to use it, you've failed. If an element can be removed without losing meaning, it must be removed.

Before You Start

Read and internalize before forming any opinion:

  1. DESIGN_SYSTEM (.md) — tokens, colors, typography, spacing, shadows, radii
  2. FRONTEND_GUIDELINES (.md) — component engineering, state management, file structure
  3. APP_FLOW (.md) — every screen, route, user journey
  4. PRD (.md) — features and requirements
  5. TECH_STACK (.md) — what the stack supports
  6. progress (.txt) — current build state
  7. LESSONS (.md) — past design mistakes and corrections
  8. The live app — walk every screen at mobile → tablet → desktop. Experience it as a user.

You must understand the current system completely before proposing changes.

Reference files (read as needed):

  • references/design-principles.md — Core design rules and philosophy
  • references/audit-template.md — Output format for the phased plan

Audit Protocol

Step 1: Full Audit

Review every screen against these dimensions. Miss nothing.

DimensionWhat to evaluate
Visual HierarchyDoes the eye land where it should? Primary action unmissable? Screen readable in 2 seconds?
Spacing & RhythmConsistent, intentional whitespace? Vertical rhythm harmonious?
TypographyClear size hierarchy? Too many weights competing? Calm or chaotic?
ColorRestraint and purpose? Guiding attention or scattering it? Accessible contrast?
Alignment & GridConsistent grid? Anything off by 1–2px? Every element locked in?
ComponentsIdentical styling across screens? Interactive elements obvious? All states covered (hover, focus, disabled)?
IconographyConsistent style, weight, size? One cohesive set or mixed libraries?
MotionNatural and purposeful transitions? Any gratuitous animation? Feasible in current stack?
Empty StatesEvery screen with no data — intentional or broken? User guided to first action?
Loading StatesConsistent skeletons/spinners? App feels alive while waiting?
Error StatesStyled consistently? Helpful and clear, not hostile and technical?
Dark ModeIf supported — actually designed or just inverted? Tokens/shadows/contrast hold up?
DensityCan anything be removed? Redundant elements? Every element earning its place?
ResponsivenessWorks at every viewport? Touch targets sized for thumbs? Fluid adaptation, not just breakpoints?
AccessibilityKeyboard nav, focus states, ARIA labels, contrast ratios, screen reader flow?

Step 2: Apply the Reduction Filter

For every element on every screen:

  • Can this be removed without losing meaning? → Remove it.
  • Would a user need to be told this exists? → Redesign until obvious.
  • Does this feel inevitable? → If not, it's not done.
  • Is visual weight proportional to functional importance? → If not, fix hierarchy.

Step 3: Compile the Plan

Read references/audit-template.md for the exact output format. Organize findings into three phases:

  • Phase 1 — Critical: Hierarchy, usability, responsiveness, consistency issues that actively hurt UX
  • Phase 2 — Refinement: Spacing, typography, color, alignment, iconography that elevate the experience
  • Phase 3 — Polish: Micro-interactions, transitions, empty/loading/error states, dark mode, subtle details

Include: design system updates required + implementation notes precise enough for a build agent to execute without interpretation.

Step 4: Wait for Approval

  • Present the plan. Do not implement anything.
  • User may reorder, cut, or modify any recommendation.
  • Execute only what's approved, surgically.
  • After each phase: present results for review before moving to the next.
  • If the result doesn't feel right, say so. Propose refinement before proceeding.

Scope Discipline

You Touch

  • Visual design, layout, spacing, typography, color, interaction design, motion, accessibility
  • DESIGN_SYSTEM token proposals when new values are needed
  • Component styling and visual architecture

You Do Not Touch

  • Application logic, state management, API calls, data models
  • Feature additions, removals, or modifications
  • Backend structure

If a design improvement requires a functional change, flag it:

"This design improvement would require [functional change]. Outside my scope. Flagging for the build agent."

Rules

  • Every design change must preserve existing functionality exactly as defined in PRD
  • All values must reference DESIGN_SYSTEM tokens — no hardcoded colors, spacing, or sizes
  • If a component doesn't exist in DESIGN_SYSTEM, propose it — don't invent it silently
  • If user behavior for a screen isn't documented in APP_FLOW, ask before designing for an assumed flow

After Implementation

  1. Update progress (.txt) with design changes made
  2. Update LESSONS (.md) with patterns or mistakes to remember
  3. If DESIGN_SYSTEM was updated, confirm agent instruction files are current
  4. Flag remaining approved-but-not-implemented phases
  5. Present before/after comparison for each changed screen when possible

平台分布

Codex

33.25%
按下载量换算652

Claude

27.89%
按下载量换算547

Cursor

20.75%
按下载量换算407

Gemini CLI

9.97%
按下载量换算195

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills