Token导航 LogoToken导航

Agent Skill · 研究检索

minoan-frontend-design(米诺斯前端设计)

用于辅助前端页面、组件、样式和交互逻辑的开发与维护。它适合让 Agent 生成或审查 React、Next.js、Vue、Tailwind、CSS 等相关代码,整理组件结构,或定位布局和性能问题。使用时需要结合项目现有设计系统、路由和构建方式,避免只生成孤立片段;涉及页面改动时,应配合本地预览和构建检查确认视觉效果。

minoan-frontend-design 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,权限为执行命令;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 46,384。

总安装 783周安装 32GitHub Stars 22下载 251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:minoan-frontend-design(米诺斯前端设计)
来源仓库:https://github.com/tdimino/claude-code-minoan
仓库路径:skills/minoan-frontend-design
安装命令:
npx skills add https://github.com/tdimino/claude-code-minoan --skill minoan-frontend-design
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/tdimino/claude-code-minoan --skill minoan-frontend-design

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。

用于辅助前端页面、组件、样式和交互逻辑的开发与维护。适合生成或审查 React、Next.js、Vue、Tailwind、CSS 等相关代码。使用时需结合项目现有设计系统、路由和构建方式,避免只生成孤立片段。

安装方式:github,通过 npx skills add 命令从 tdimino/claude-code-minoan 仓库添加。适用于 Codex、Claude、Cursor、Gemini CLI 等宿主环境。

来源仓库内路径:skills/minoan-frontend-design

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/tdimino/claude-code-minoan --skill minoan-fron…;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

    它需要执行命令,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

minoan-frontend-design 怎么安装?

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

minoan-frontend-design 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。

minoan-frontend-design 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/tdimino/claude-code-minoan --skill minoan-frontend-design 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

minoan-frontend-design 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。

minoan-frontend-design 用的人多吗?

公开数据是累计安装 783 次,近一周安装 32 次,下载 251 次,GitHub 22 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。

minoan-frontend-design 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

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 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

  • 不能接受 Skill 在你机器上执行命令的人。
  • 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
minoan-frontend-design

米诺斯前端设计 · 本页

78322执行命令未明确通过

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

Build frontend interfaces that are unmistakably *designed*, not generated. Every interface should feel like a specific designer made specific choices for a specific context.

Creative Direction

Name the conceptual direction before coding—"Bloomberg Terminal," "medical journal," "zine collage," "VHS dreamworld." The name anchors every subsequent decision.

Root every aesthetic choice in the subject itself. A jazz club's website should feel like smoke and brass. A children's reading app should feel like opening a pop-up book. A fintech dashboard should feel like a control room, not a toy. The design should be so fitted to its context that transplanting the aesthetic to a different project would feel wrong.

Commit to a singular direction: brutally minimal, maximalist chaos, luxury/refined, lo-fi/zine, dark and moody, editorial/magazine, brutalist/raw, retro-futuristic, art deco/geometric, cinematic/widescreen, data-dense/scientific, or something entirely unnamed. The final design should feel like nothing else.

Every design needs one element so distinctive someone would describe it to a friend. Find that element and let everything else serve it.

Typography

Typography carries the design's voice—it should have weight you can feel. Default fonts signal default thinking: skip Arial, Inter, Roboto, Space Grotesk, system stacks. Display type should provoke. Body text should be legible and refined. Pair them like actors in a scene—tension creates character. Work the full typographic range: dramatic size jumps, weight contrasts, uppercase for authority, generous letter-spacing for elegance. Control hierarchy through weight and color, not just scale.

Massive editorial typography—oversized display type at 6xl-9xl, ghost numbers behind content, typographic strikethrough—signals authority. At least one typographic moment per design should make a statement, not just organize information.

Color

Color sets the room's temperature. Palettes should take a clear position: bold and saturated, moody and restrained, or high-contrast and minimal—never timid, never non-committal. Lead with a dominant color, punctuate with sharp accents. A single unexpected accent color does more than a balanced five-color system. Never pure black—pure black creates harsh contrast that flattens depth. Use rich off-blacks (zinc-950, deep navy). Tint shadows to the background hue. Warm editorial palettes (cream, parchment, amber on charcoal) are as bold as dark mode—resist defaulting to dark when the context calls for warmth. When dark themes fit, commit fully: singular accent color (acid yellow, bioluminescent cyan, deep rose) against rich off-blacks.

Motion

Prioritize CSS-only animations—content hidden behind JavaScript class toggles is invisible in screenshots and static renders. Start visible, animate from there. Staggered page-load reveals create more delight than scattered micro-interactions. Scroll-triggered animations and hover states should surprise.

Spatial Composition

Unexpected layouts. Asymmetry. Overlap and z-depth. Diagonal flow. Grid-breaking elements. Dramatic scale jumps. Full-bleed moments. Split-screen compositions. Bento grids with varied tile sizes. Generous negative space OR controlled density—commit to one, never the uncommitted middle. Centered hero sections are the default—break that default with left-aligned text, asymmetric splits, or editorial whitespace. Use cards only when elevation communicates hierarchy, not as a generic container for everything.

For dashboards and data-dense interfaces, maximize information above the fold. Pack metrics, charts, and controls into dense grids separated by thin borders and negative space—not card containers. Use monospace for all numbers. Treat the dashboard as a command center, not a gallery.

Atmosphere & Texture

Create depth and mood rather than defaulting to flat solid colors. The background is a canvas, not a wall. Apply contextual effects: gradient meshes, noise and grain overlays, geometric patterns, glassmorphism with inner refraction borders, dramatic or soft shadows, parallax depth, clip-path shapes, print-inspired textures (halftone, duotone, risograph), knockout typography, hand-drawn SVG accents, generative pattern backgrounds.

Generate custom SVG illustrations fitted to the context—never placeholder images. A ramen shop gets a hand-drawn bowl. An ocean nonprofit gets bioluminescent particles. The illustration is the design, not decoration.

What to Reject

Never use generic AI aesthetics: overused fonts (Inter, Roboto, Arial, Space Grotesk, system stacks), cliched color schemes (purple gradients on white, neon outer glows), predictable centered layouts, the generic three-equal-cards feature row, excessive gradient text, generic placeholder content (John Doe, 99.99%, "Elevate your workflow"), emoji or SVG-egg avatar placeholders, hero sections that push core content below the fold, or incomplete implementations. Instead: distinctive fonts that take a position. Bold, committed palettes. Asymmetric layouts that surprise. Bespoke details rooted in context. Realistic content with creative names and organic data (47.2%, not 50%). When photos are requested, use image services or generate meaningful visual content—never empty containers. Every feature requested, every choice deliberate.

Resist the first satisfying idea. The model's default output is the average of its training data—genuine distinction means pushing past the obvious.

Craft

Match implementation complexity to the aesthetic vision. Maximalist designs demand elaborate code. Minimalist designs demand obsessive restraint. Both demand pixel-level attention to spacing, alignment, and consistency. Deliver every feature the prompt requests—completeness is the floor, not the ceiling. A complete AND bold implementation beats both a partial masterpiece and a complete bore.

Claude is capable of extraordinary, award-worthy creative work. The person who sees this interface should feel something they didn't expect to feel. Don't hold back—show what's truly possible when every aesthetic choice is intentional, every detail is refined, and the vision is executed without compromise.

Post-Build QA Workflow

After building, run these refinement passes in order:

  1. Audit: Check against references/anti-patterns.md + domain-specific impeccable refs below. Score 5 dimensions (a11y, performance, responsive, theming, anti-patterns) 0-4 each. Tag issues P0-P3 per references/heuristics-scoring.md.
  2. Critique: UX review using references/heuristics-scoring.md (Nielsen's 10 heuristics, score 0-4 each, /40). Assess cognitive load via references/cognitive-load.md (8-item checklist). Test with 2-3 personas from references/personas.md. Run the AI Slop Test: if someone said "AI made this," would they believe it immediately? If yes, fix it.
  3. Normalize: Align with project design system if .design-context.md or DESIGN.md exists in project root. Treat their tokens (colors, fonts, spacing) as materials, not a recipe—creative direction from above still leads.
  4. Harden: Test edge cases -- long text overflow, RTL, empty states, error states, i18n, keyboard navigation.
  5. Polish: Final pass -- alignment, spacing, interaction states, transitions, micro-copy. Every pixel intentional.

References

Consult references/ on demand:

Engineering: vercel-web-interface-guidelines.md, design-system-checklist.md. Techniques: creative-arsenal.md, design-dials.md, editorial-patterns.md (eval-tested), astryx-hero.md (depth-parallax editorial hero + fluid-dom sibling links). Shader cross-ref: ~/.claude/skills/rocaille-shader/references/fluid-dom.md (pointer-reactive fluid over DOM sections). Stripe (light-mode): stripe-design-tokens.md, stripe-component-patterns.md, stripe-signature-techniques.md. Mintlify (dark-first docs): mintlify-design-tokens.md, mintlify-component-patterns.md, mintlify-signature-techniques.md. Anti-patterns: anti-patterns.md (condensed checklist from Impeccable v1.6.0). Critique & Scoring: heuristics-scoring.md (Nielsen's 10, P0-P3 severity), cognitive-load.md (8-item checklist, working memory rule), personas.md (5 archetypes + project-specific). Design system exemplar: linear-design-md.md (DESIGN.md format—tokens, components, responsive rules from Linear). Advanced techniques: sunny-mode-technique.md (@property palette transitions, blend-mode overlays, physics Easter eggs). Impeccable domain refs (code examples + implementation patterns): impeccable-typography.md (font loading, OpenType, scales), impeccable-color-contrast.md (OKLCH, dark mode tokens), impeccable-spatial.md (container queries, optical alignment), impeccable-motion.md (cubic-bezier values, stagger formula, reduced-motion), impeccable-interaction.md (Popover API, roving tabindex, dialog), impeccable-responsive.md (pointer/hover queries, safe areas, srcset), impeccable-ux-writing.md (error templates, voice matrix). Color science (from meodai/skill.color-expert): color-science-deep.md (color spaces decision matrix, HSL limitations, chroma vs saturation, character-first harmony, accessibility statistics, pigment mixing, naming systems), color-tools-palette.md (palette generation algorithms, code libraries, analysis/linting, CSS Color 4/5, semantic token architecture, generative art approaches). Component research: invoke component-gallery.

平台分布

Claude

33.8%
按下载量换算85

Codex

33%
按下载量换算83

Cursor

17.74%
按下载量换算45

Gemini CLI

10.21%
按下载量换算26

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/tdimino/claude-code-minoan --skill minoan-frontend-design 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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