Token导航 LogoToken导航

Agent Skill · 前端设计

enhance-prompt(增强提示)

用于辅助提示词、系统指令、Agent 行为约束和工作流模板的整理。它适合让 Agent 规范任务边界、统一输出格式、拆分操作步骤或优化提示词可复用性。使用时需要保留真实业务约束,不要把示例当硬规则;涉及自动执行、外部工具或高风险操作时,应在提示词中明确确认步骤、权限边界和失败处理方式。

enhance-prompt 是前端设计类 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,第三方扫描有提醒项,按安装量在前端设计类的 10,216 个 Skill 里排第 5,673。

总安装 624周安装 25GitHub Stars 25下载 202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:enhance-prompt(增强提示)
来源仓库:https://github.com/oimiragieo/agent-studio
仓库路径:skills/enhance-prompt
安装命令:
npx skills add https://github.com/oimiragieo/agent-studio --skill enhance-prompt
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/oimiragieo/agent-studio --skill enhance-prompt

这个 Skill 是做什么的

前端设计类 Skill 关注页面结构、视觉规范和组件写法,输出通常是可运行的界面代码。

enhance-prompt 用于辅助提示词、系统指令、Agent 行为约束和工作流模板的整理。适用于提示词优化、工作流设计和系统指令制定等场景。通过 GitHub 仓库安装,使用 npx skills add 命令添加技能。

使用时需要保留真实业务约束,不要把示例当硬规则;涉及自动执行、外部工具或高风险操作时,应在提示词中明确确认步骤、权限边界和失败处理方式。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/enhance-prompt

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/oimiragieo/agent-studio --skill enhance-prompt;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

enhance-prompt 怎么安装?

在终端执行 npx skills add https://github.com/oimiragieo/agent-studio --skill enhance-prompt。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

enhance-prompt 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。

enhance-prompt 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。

enhance-prompt 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。

enhance-prompt 用的人多吗?

公开数据是累计安装 624 次,近一周安装 25 次,下载 202 次,GitHub 25 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。

enhance-prompt 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

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。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。
  • 需要处理前端设计类任务、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。

前端设计类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
frontend-design

前端设计 · Codex、Claude

8,823,360126,072只读MIT通过
vercel-composition-patterns

Vercel 组合模式 · Codex、Claude

3,858,81625,893需要联网MIT通过
vercel-react-native-skills

Vercel React native skills 前端 · Codex、Claude

2,468,66425,878只读MIT通过
adapt

adapt 搜索 · Codex、Claude

2,027,04023,125操作浏览器未明确通过
bolder

bolder 分析 · Codex、Claude

2,004,28823,130只读未明确通过
enhance-prompt

增强提示 · 本页

62425需要联网未明确有提醒

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

Enhance Prompt

Transform vague UI and feature requests into structured, optimized prompts with design system awareness. Based on the Google Labs Stitch Skills prompt enhancement pipeline.

When to Apply

Use this skill when:

  • A user request describes a UI feature vaguely ("make a nice login page")
  • You need to generate implementation prompts from loose requirements
  • Translating design mockups or wireframes into actionable specs
  • Integrating design system tokens into component specifications
  • Scaffolding page layouts from high-level descriptions

4-Step Enhancement Pipeline

Step 1: Assess Intent

Analyze the raw request to identify:

  • Core objective: What is the user trying to build?
  • Implicit requirements: What is assumed but not stated?
  • Target audience: Who will use this UI?
  • Interaction patterns: What user flows are implied?

Example:

Raw: "Add a settings page"
Assessment:
  - Core: User preferences management interface
  - Implicit: Navigation integration, form validation, save/cancel actions
  - Audience: Authenticated users
  - Interactions: View current settings, modify, save, receive confirmation

Step 2: Check DESIGN.md

Look for a DESIGN.md or equivalent design system file in the project root or docs directory. Extract:

  • Color tokens: Primary, secondary, accent, semantic colors
  • Typography scale: Font families, sizes, weights, line heights
  • Spacing system: Base unit, scale (4px, 8px, 12px, 16px, etc.)
  • Component inventory: Available pre-built components
  • Layout patterns: Grid system, breakpoints, container widths

If no design system file exists, generate reasonable defaults based on the project's existing patterns (check tailwind.config.*, theme.*, or CSS custom properties).

Lookup order:

  1. DESIGN.md in project root
  2. docs/design-system.md
  3. tailwind.config.ts / tailwind.config.js theme section
  4. CSS custom properties in global styles
  5. shadcn/ui theme configuration (if present)

Step 3: Apply Enhancements

Transform the raw request by applying these enhancement layers:

3a. Add Specificity

Replace vague terms with specific component names and patterns:

Vague TermSpecific Replacement
"nice"Clean layout with consistent spacing, proper visual hierarchy
"modern"Card-based layout, rounded corners, subtle shadows, micro-interactions
"responsive"Mobile-first grid (1-col mobile, 2-col tablet, 3-col desktop)
"fast"Skeleton loading states, optimistic updates, prefetched data
"user-friendly"Clear labels, inline validation, helpful error messages, focus management
"beautiful"Design-system-compliant colors, balanced whitespace, typography scale
"simple"Single-column layout, minimal navigation, progressive disclosure
"interactive"Hover states, transitions, keyboard navigation, ARIA attributes
3b. Inject Design Tokens

Insert design system values into the specification:

Enhanced: "Use the primary color (--color-primary / colors.primary.500)
for the submit button. Apply spacing-4 (16px) between form fields.
Use the heading-2 typography preset for section titles."
3c. Scaffold Layout Structure

Generate a structural layout for the page/component:

Page Layout:
  Header: Breadcrumb navigation + page title
  Body:
    Sidebar (desktop only): Section navigation
    Main Content:
      Section 1: [Primary feature area]
      Section 2: [Secondary feature area]
      Section 3: [Related actions]
  Footer: Save/Cancel actions (sticky on mobile)
3d. Add Interaction Specifications

Specify loading states, error states, empty states, and transitions:

  • Loading: Skeleton placeholder matching final layout
  • Error: Inline error messages with retry action
  • Empty: Helpful illustration + primary action CTA
  • Success: Toast notification + state update
  • Transitions: 200ms ease-in-out for element appearance

Step 4: Format Output

Produce the enhanced prompt in this structured format:

## Enhanced Specification: [Feature Name]

### Objective

[1-2 sentences describing the precise goal]

### Design System Integration

- Colors: [specific tokens]
- Typography: [specific presets]
- Spacing: [specific scale values]
- Components: [reusable components to use]

### Layout Structure

[ASCII or markdown layout diagram]

### Component Breakdown

1. [Component name] - [purpose] - [key props/state]
2. [Component name] - [purpose] - [key props/state]

### Interaction States

- Loading: [description]
- Error: [description]
- Empty: [description]
- Success: [description]

### Accessibility Requirements

- Keyboard navigation: [specifics]
- Screen reader: [ARIA labels/roles]
- Focus management: [focus trap/restore behavior]

### Responsive Behavior

- Mobile (< 640px): [layout]
- Tablet (640-1024px): [layout]
- Desktop (> 1024px): [layout]

Complete Example

Raw Request: "Build a dashboard for our analytics"

Enhanced Output:

## Enhanced Specification: Analytics Dashboard

### Objective

Build a data visualization dashboard displaying key business metrics
with real-time updates, date range filtering, and exportable reports.

### Design System Integration

- Colors: --color-primary for active states, --color-success/warning/danger
  for metric status indicators
- Typography: heading-1 for page title, heading-3 for card titles,
  body-sm for metric labels
- Spacing: gap-6 between metric cards, padding-4 inside cards
- Components: Card, Select (date picker), Button, Badge, Skeleton

### Layout Structure

[Nav]
[Page Title + Date Range Selector + Export Button]
[KPI Card] [KPI Card] [KPI Card] [KPI Card] <- 4-col grid
[Line Chart (2/3 width)] [Pie Chart (1/3 width)] <- 2-col split
[Data Table (full width)] <- sortable, paginated

### Component Breakdown

1. KPICard - Displays single metric with trend indicator - { label, value, change, status }
2. DateRangeSelector - Preset ranges + custom date picker - { range, onRangeChange }
3. TrendChart - Line/area chart for time series data - { data, dateRange, metric }
4. DistributionChart - Pie/donut for categorical data - { data, metric }
5. MetricsTable - Sortable, paginated data table - { columns, data, sortBy, page }

### Interaction States

- Loading: Skeleton cards (pulse animation) matching final dimensions
- Error: Error banner with retry button, individual chart error boundaries
- Empty: "No data for selected range" with suggestion to expand range
- Success: Smooth number transitions (count-up animation) on data refresh

### Accessibility Requirements

- Keyboard navigation: Tab through cards, charts have data table alternatives
- Screen reader: Chart descriptions via aria-label, live region for metric updates
- Focus management: Date picker focus trap, return focus on close

### Responsive Behavior

- Mobile (< 640px): Single column, KPI cards stack vertically, charts full width
- Tablet (640-1024px): 2-col KPI grid, charts stack vertically
- Desktop (> 1024px): 4-col KPI grid, side-by-side charts

Anti-Patterns

  • Do NOT skip Step 2 (design system check) -- prompts without design tokens produce inconsistent UI
  • Do NOT leave vague terms unresolved -- "nice" is not a specification
  • Do NOT generate prompts without interaction states -- every view has loading, error, and empty states
  • Do NOT ignore accessibility -- every enhanced prompt must include keyboard and screen reader specs
  • Do NOT hardcode pixel values -- always reference design system tokens or relative units

Iron Laws

  1. ALWAYS analyze the original prompt for ambiguities and implicit assumptions before enhancing — never enhance a prompt whose scope you haven't verified with the caller.
  2. NEVER add requirements that weren't implicit or explicit in the original — enhancement clarifies and structures; it does not invent scope.
  3. ALWAYS preserve the original intent — an enhanced prompt that redirects to a different goal is a rewrite, not an enhancement.
  4. NEVER produce an enhanced prompt longer than necessary to resolve ambiguities — verbosity in prompts reduces AI response quality; keep it focused.
  5. ALWAYS include success criteria in the enhanced prompt — a prompt without acceptance criteria cannot be evaluated as done or not done.

Memory Protocol (MANDATORY)

After using this skill, record learnings using MemoryRecord:

MemoryRecord({
  type: 'pattern',
  text: 'enhance-prompt: [describe what design system pattern or prompt structure worked well]',
  area: 'frontend',
});

Write decisions to .claude/context/memory/decisions.md when choosing between design system approaches. Write issues to .claude/context/memory/issues.md when DESIGN.md is absent and fallback patterns were used.

平台分布

Codex

37.76%
按下载量换算76

Claude

31.66%
按下载量换算64

Cursor

17.15%
按下载量换算35

Gemini CLI

9.15%
按下载量换算18

安全审计

Gen Agent Trust Hub

可疑

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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