Token导航 LogoToken导航

Agent Skill · 研究检索

shadcn-component-discovery(shadcn/ui component discovery 搜索)

shadcn-component-discovery 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

shadcn-component-discovery 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,权限为调用外部服务;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 26,529。

总安装 3,550周安装 145GitHub Stars 4下载 1,137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:shadcn-component-discovery(shadcn/ui component discovery 搜索)
来源仓库:https://github.com/mattbx/shadcn-skills
仓库路径:skills/shadcn-component-discovery
安装命令:
npx skills add https://github.com/mattbx/shadcn-skills --skill shadcn-component-discovery
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/mattbx/shadcn-skills --skill shadcn-component-discovery

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。

shadcn-component-discovery 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。适用于从 shadcn/ui 生态中发现可用组件或第三方扩展,支持快速集成。

通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 确认具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写操作。可结合来源仓库和 SKILL.md 继续核验功能细节,确保与项目需求匹配。

来源仓库内路径:skills/shadcn-component-discovery

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/mattbx/shadcn-skills --skill shadcn-component-…;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

    它需要调用外部服务,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

shadcn-component-discovery 怎么安装?

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

shadcn-component-discovery 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。

shadcn-component-discovery 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

shadcn-component-discovery 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。

shadcn-component-discovery 用的人多吗?

公开数据是累计安装 3,550 次,近一周安装 145 次,下载 1,137 次,GitHub 4 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。

shadcn-component-discovery 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
shadcn-component-discovery

shadcn/ui component discovery 搜索 · 本页

3,5504调用外部服务未明确有提醒

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

shadcn Component Discovery

Surface existing components from across the shadcn ecosystem — including registries the user hasn't configured yet — before spending effort on custom implementations.

How This Complements the Official shadcn Skill

The official shadcn skill (shipped with CLI v4) searches the registries declared in the user's components.json. That's the right tool once the user has decided which registries to use.

This skill fills the *upstream* gap: ecosystem-wide awareness. It helps the user find and choose registries they may not yet have configured — Magic UI for animations, Aceternity for hero sections, AI Elements for chat interfaces, ElevenLabs UI for voice components, and so on. Use this skill for "what exists out there"; let the official skill handle "install this thing I know about."

When both skills apply, this one runs first (discovery), then the official skill takes over for installation and project-aware work.

Core Principle

Search before building. Most UI needs have been solved in the shadcn ecosystem. The cost of a 10-second registry check is almost always less than custom-building something that already exists.

When to Trigger

Proactive triggers (activate before writing any component code):

  • User asks to build a table, form, modal, sidebar, dashboard, auth page, landing section, animation, carousel, chart, or any standard UI pattern
  • User describes a feature that implies UI (e.g. "add user accounts management" → table; "let users pick a date range" → date picker)
  • User mentions building something "like X" where X is a common pattern

Explicit triggers (user directly requests search):

  • "Find a component for..."
  • "Is there a shadcn component for..."
  • "Search registries for..."
  • "What exists for..."
  • "Any good [category] components?"
  • Direct mentions of specific registries (@animate-ui, Magic UI, Aceternity, Tailark, etc.)

Workflow

Step 1: Clarify the Need

Before searching, confirm:

  • Functionality: What must it do? (e.g. sortable + filterable + paginated table)
  • Style: Animated, minimal, accessible-first, dense, brutalist?
  • Constraints: Must support keyboard nav? Drag-drop? Mobile-first? SSR?

A two-sentence clarification beats five searches with the wrong query.

Step 2: Search

Three paths depending on tool availability. The official shadcn MCP is strongly preferred when the user has a shadcn project — it returns real results with code examples.

Path A: Official shadcn MCP (preferred)

If the shadcn MCP server is running (configured via npx shadcn@latest mcp init or equivalent), use its tools directly:

1. mcp__shadcn__search_items_in_registries
   - registries: ["@shadcn", "@blocks", "@animate-ui", "@reui", "@diceui"]
   - query: "<search term>"
   - limit: 10

2. For promising results, get full details:
   mcp__shadcn__view_items_in_registries
   - items: ["@registry/component-name", "@registry/other"]

3. For implementation examples:
   mcp__shadcn__get_item_examples_from_registries
   - query: "<component>-demo"

4. For the install command:
   mcp__shadcn__get_add_command_for_items
   - items: ["@registry/chosen-component"]

The MCP only searches registries configured in the user's components.json. If the user is looking for something specialized, suggest adding a relevant registry to components.json first — see references/registries.md for recommendations and configuration snippets.

Path B: shadcn CLI (no MCP, but shadcn project)

If there's a components.json but no MCP is configured, use the CLI commands that shipped with CLI v3:

# Search across configured registries
npx shadcn@latest search "<term>"

# Preview a component before installing
npx shadcn@latest view @registry/component-name

# List everything available in a registry
npx shadcn@latest list @registry

Suggest adding the MCP for better agent integration: npx shadcn@latest mcp init.

Path C: No project context

If there's no shadcn project yet, or the user is just exploring:

  • Recommend registries from references/registries.md based on their described need
  • Link directly to the registry's website so they can browse
  • Offer to help set up a shadcn project (npx shadcn@latest init) once they've picked direction

Step 3: Present Findings

Match the response format to the context. Keep it scannable — the user is about to make a decision, not read an essay.

Proactive check (during build flow)

When the user didn't explicitly ask to search but you searched anyway:

Before building custom, I found existing options:

1. **@registry/component-name** — [one-line description]
2. **@registry/alternative** — [one-line description]

Want me to install one, or build custom?

Keep it under five lines. The user is in flow; don't derail them.

Explicit search (user asked)
## Results for "<query>"

**Top matches:**

### @registry/component-name ⭐
[What it does in one sentence]
- Fits because: [specific to their need]
- Install: `npx shadcn@latest add @registry/component-name`

### @registry/alternative
[What it does in one sentence]
- Fits because: [specific to their need]
- Install: `npx shadcn@latest add @registry/alternative`

---
Install [1] or [2], see more results, view code example, or build custom?
Comparison (multiple good options, choice matters)

Use a short table when 3+ options are genuinely viable and trade off differently. Follow with a recommendation and the install command for the recommended one. Don't table-ify everything — it's heavier than plain prose and should earn its place.

Step 4: Execute the Choice

  • Install chosen: Run the add command, then customize via props / composition as needed
  • See more: Return additional matches, continue pagination
  • View code: Fetch example via get_item_examples_from_registries or the registry's demo URL
  • Build custom: Proceed — but reference the closest existing component for composition patterns (CVA, cn(), data-slot, semantic tokens)

Search Strategy

Effective Query Terms

Short, concrete nouns work best. Keep queries to 1–3 words.

Looking for…Try searching…
Data displaytable, data-grid, list
User inputform, input, field, select, combobox
Navigationsidebar, nav, menu, tabs, breadcrumb
Feedbacktoast, sonner, alert, notification
Overlaysdialog, modal, sheet, popover, drawer
Mediacarousel, gallery, image
Animationanimate, motion, transition
Layoutcard, section, hero, grid, stack
Date/timecalendar, date-picker, date-range

Registry Specialties (Quick Pick)

NeedCheck first
Core UI primitives@shadcn
Page sections, pre-built blocks@blocks, Tailark, HextaUI
Data grids, advanced tables@reui
Animated components@animate-ui, Magic UI, Aceternity UI
AI / chat interfacesAI Elements, assistant-ui
Accessibility-first (React Aria)@diceui, JollyUI, Intent UI
Voice / audioElevenLabs UI
Marketing / landing pagesTailark, Eldora UI
Style variants (brutalist, retro, glass)See references/registries.md

For the full catalog including URLs, configuration snippets, and registry specialties, see references/registries.md.

Response Requirements

Every discovery response should surface:

  1. What was searched — the query used
  2. Top 2–3 recommendations — registry-qualified name, one-line description, why it fits
  3. Install command — ready to copy
  4. A recommendation — which to pick, when there's a clear fit
  5. Next-step options — install, view code, see more, build custom

Skip anything beyond this unless the user asks. Brevity respects flow.

Examples

Proactive check

User: "Add a table showing user accounts with sorting"

Response:

Before building custom, I found existing options: 1. @reui/data-grid-table — Full-featured grid with sorting, filtering, pagination 2. @shadcn/table — Unstyled primitives (pair with TanStack Table for sorting) @reui/data-grid-table looks like the fit — install it?

Explicit search

User: "Find me an animated accordion"

Response:

## Results for "animated accordion" ### @animate-ui/components-base-accordion ⭐ Accordion with smooth enter/exit transitions built on Motion. - Fits because: Animation is built-in, accessible, keyboard nav included - Install: npx shadcn@latest add @animate-ui/components-base-accordion ### @shadcn/accordion Standard Radix-based accordion, unanimated by default. - Fits because: Lighter dependency footprint; you add your own motion - Install: npx shadcn@latest add @shadcn/accordion --- Install animated [1], install plain [2], or see more?

No good match

User: "Find a component for visualizing git commit lineage"

Response:

## Results for "git commit lineage" No directly matching components in the shadcn ecosystem. Closest adjacencies: - @reui/tree — tree visualizations (could adapt) - @shadcn/chart — generic charting (Recharts-based) This is a custom build. Want me to sketch an approach using one of these as a foundation, or look at non-shadcn libraries like @gitgraph/react?

Best Practices

Do:

  • Search before writing component code — always
  • Surface 2–3 options, not 10
  • Explain *why* each fits the specific need
  • Include the install command, not just the name
  • Recommend when the choice is clear

Don't:

  • Skip searching because "it's faster to build" (it usually isn't)
  • Dump the full search-result list — curate
  • Present registries the user hasn't configured without mentioning the components.json step
  • Recommend unmaintained or abandoned registries (check references/registries.md for current status)

Resources

平台分布

Codex

35.52%
按下载量换算404

Claude

28.64%
按下载量换算326

Cursor

18.18%
按下载量换算207

Gemini CLI

8.63%
按下载量换算98

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

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