Token导航 LogoToken导航

Agent Skill · 研究检索

vibe-deck(氛围甲板)

vibe-deck 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。

vibe-deck 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 46,976。

总安装 753周安装 32GitHub Stars 3下载 264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:vibe-deck(氛围甲板)
来源仓库:https://github.com/aiden0z/skills
仓库路径:skills/vibe-deck
安装命令:
npx skills add https://github.com/aiden0z/skills --skill vibe-deck
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/aiden0z/skills --skill vibe-deck

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。

用于关键词检索与信息筛选,快速定位候选结果。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。适合在任务场景下根据线索查找相关文档或资源。通过 Agent 执行搜索并返回结构化摘要,提升信息获取效率。安装前应确认是否会触发联网或文件操作,避免越权访问。

vibe-deck 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/vibe-deck

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/aiden0z/skills --skill vibe-deck;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

vibe-deck 怎么安装?

在终端执行 npx skills add https://github.com/aiden0z/skills --skill vibe-deck。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

vibe-deck 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。

vibe-deck 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。

vibe-deck 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。

vibe-deck 用的人多吗?

公开数据是累计安装 753 次,近一周安装 32 次,下载 264 次,GitHub 3 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。

vibe-deck 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
vibe-deck

氛围甲板 · 本页

7533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

VibeDeck

Step 0: Environment Check

Before anything else, verify that Node.js and npm are available. Try these commands (use whichever works on the current platform):

node --version    # or: where.exe node (Windows)
npm --version     # or: where.exe npm (Windows)
  • Both exist and Node.js ≥ 20 → continue silently (do not mention the check to the user).
  • Missing or version too low → do NOT proceed. Instead:

1. Run uname -s and uname -m (or systeminfo on Windows) to detect the user's OS and architecture. 2. If you have web search capability, search for the latest recommended way to install Node.js on the user's platform, and provide the specific commands. 3. If you cannot search, show this message in the user's language: 中文: VibeDeck 需要 Node.js 20 或更高版本。请访问 https://nodejs.org 下载并安装适合你操作系统的版本,安装完成后重新运行。 English: VibeDeck requires Node.js 20 or later. Please visit https://nodejs.org to download and install the version for your OS, then try again. 4. Stop here and wait for the user to install before continuing.

Step 1: Check Project

Check if slide-kit.config.js exists in the current directory.

  • Exists → this is an existing slide-kit project. Go to Building Slides.
  • Does not exist → no project yet. Go to Creating a New Deck, then continue to Building Slides if the user also described content.

Creating a New Deck

1. Ask for title

Only ask for the project title (required). Show defaults for everything else:

I'll set up the project with these defaults unless you want changes: - Directory: ./<title-slug>/ - Theme: corporate-blue (also available: minimal) - Logo: built-in SlideKit logo - Presenter: none (can set later in config) - Password: none Let me know if you want to change any of these, or I'll proceed.

See theme-presets.md for theme details.

2. Scaffold the project

cp -r <this-skill-dir>/template/ <target-dir>
cd <target-dir>
git init

No network required — the template is bundled in this skill.

3. Configure

  • slide-kit.config.js: set title, theme, logo, presenter as needed

- Custom primary color: use overrides.colors.primary — variants auto-derived - Custom logo: copy to public/, reference as /filename.svg

  • index.html: set <title>
  • src/App.jsx: remove PasswordGate if no password

4. Generate agent instructions

Read instruction-template.md, replace placeholders, write to both CLAUDE.md and AGENTS.md.

5. Install and start

npm install
npm run dev

Report: project location, theme, dev server URL.


Building Slides

1. Read context (do this BEFORE planning)

Read these three files first — skipping this step leads to duplicated content, wrong theme colors, or missed conventions:

  • CLAUDE.md (or AGENTS.md) — theme, conventions, data sources
  • src/App.jsx — current slide list and ordering (what already exists)
  • slide-kit.config.js — active theme and presenter info

2. Understand the request

  • What content should this slide show?
  • Is there data involved? Where does it come from?
  • Narrative continuity: new slide should follow logically from the previous one
  • No duplication: if a data point is already shown elsewhere, present from a different angle
  • Unit consistency: confirm the same units as existing slides ($K vs $M, etc.)
  • Content density: Check content-rules.md for minimum content density standards — slides with sparse content (< 10 visual elements, single-line card descriptions) are not acceptable.

3. Layout Decision Engine

Don't guess the layout — derive it systematically. Read layout-engine.md and follow this sequence:

  1. Intent Classification — what type of slide is this? (data / concept / showcase / reference / tool / narrative). Different types have different density tolerances — a showcase slide can be dense on purpose, a data slide cannot.
  2. Content Inventory — count metrics, charts, text blocks, table rows, list items
  3. Single Takeaway Check — state the slide's ONE takeaway. For concept slides, "understand X" can naturally combine definition + components + comparison — that's one takeaway, not three.
  4. Semantic Match — use the content-to-layout mapping table to pick the layout
  5. Capacity Check — verify content fits within the layout's limits (which vary by intent type). If not: adapt or split — but don't split showcase/reference slides just for being dense.
  6. Composite Layout — if content is mixed (e.g., chart + metrics), use a composite pattern from layout-engine.md Phase 4

Available layouts — see layout-templates.md for code:

LayoutBest for
FullChartSingle large chart + KeyMessage
SplitViewSide-by-side comparison
MetricGridDashboard with N metric cards
ComparisonViewBefore/after, similar-depth comparison
DataTableTool comparison, feature matrix, pricing
TimelineFlowRoadmap, milestones, phased plans
CardGrid4–6 feature items with icon + text
CardRow2–4 items in a single row
ConceptSlideExplanation / text-heavy, no data
CompositeMixed content — see layout-engine.md Phase 4

Built-in slide templates (import from src/slides/):

TemplateUse case
SlideCoverCover page (reads config.title automatically)
SlideDividerChapter divider/transition
SlideAgendaTable of contents
SlideThankYouClosing/Q&A page

4. Build the slide

5. Register (checklist)

import SlideXxx from './slides/SlideXxx'

// In <Deck>:
<Slide title="Page Title"><SlideXxx /></Slide>

Before moving on, verify:

  • import added at top of App.jsx
  • <Slide title="..."> — title is set (powers navigator, do NOT omit)
  • Placed in correct position within <Deck> (narrative order)
  • footnote prop set if slide uses external data

6. Storyline review (AFTER adding slides)

Every time new slides are added, re-read the full <Deck> in App.jsx and verify the overall narrative:

  • Logical flow: Does each slide follow naturally from the previous one? Would the audience be confused by the transition?
  • Progressive disclosure: Concepts introduced before they're referenced? (e.g., "What is a Skill?" must come before "Recommended Skills")
  • Audience-appropriate order: For non-technical audiences, order from simple → complex, familiar → new (e.g., VS Code/Copilot before CLI tools)
  • No orphan slides: Every slide belongs to a clear section. If a slide doesn't fit any section, it may not belong in this deck
  • Section balance: Each section should have 2-5 slides. A section with 1 slide feels incomplete; a section with 8+ slides needs splitting
  • Title scan test: Read ONLY the slide titles in order — does the story make sense from titles alone? If not, rename titles to be more descriptive

7. Layout verification (AFTER building)

After building each slide, visually verify the layout or use the layout checklist:

  • No bottom whitespace: Content fills ≥ 80% of slide height
  • No bunched content: Use justify-center or justify-between, not justify-start
  • Consistent card heights: Grid cards use h-full to fill their cell
  • Card internal alignment: Title at top (shrink-0), body in flex-1 justify-center — see content-rules.md > Card Layout Alignment for the pattern and common mistakes
  • One card per grid cell: No label + card wrappers; mockups go directly in cells
  • Balanced columns: Unequal content? Split into 3 cols, stack vertically, or center-fill
  • Text hierarchy: 28px title → 10px section label → 14px card title → 12px body → 10px aux
  • Section spacing: mt-2 or mt-3 between sections, not mt-auto
  • Card gap ratio: Gap between cards ≈ 1.6× card internal padding (e.g., py-2.5 → gap-4). See content-rules.md > Card Spacing
  • No emoji: Use lucide-react icons, not emoji characters. Store as component refs in data arrays (icon: Mail), render as <item.icon size={14} />

8. Data extraction (if needed)

  • Use scripts/extract-xlsx.js to inspect Excel structure
  • Write a custom extraction script, verify totals match
  • Save to src/data/<name>.js
  • NEVER fabricate data

Modifying Existing Slides

When the user asks to change, reorder, or remove slides in an existing deck:

1. Read context first

Same as Building Slides — read CLAUDE.md, App.jsx, and slide-kit.config.js before touching anything.

2. Editing a slide

  • Open the slide component file (e.g., src/slides/SlideRevenue.jsx)
  • Make the requested changes while preserving the existing layout pattern
  • If the change involves data, update src/data/ files and verify totals still match
  • After editing, run the Layout verification checklist (Step 7 from Building Slides)

3. Reordering slides

  • Reorder the <Slide> entries in App.jsx — this is the only file that controls order
  • After reordering, run the post-edit verification below

4. Removing slides

  • Remove the <Slide> entry from App.jsx
  • Check if any other slide references data or concepts introduced by the removed slide — if so, move that context to an earlier slide
  • Remove the slide component file if no longer imported
  • Run the post-edit verification below

5. Replacing a slide's layout

If the user wants to change a slide from one layout to another (e.g., FullChart → SplitView):

  • Read layout-templates.md for the new layout's code template
  • Rewrite the component using the new layout, preserving the data and keyMessage
  • Verify the chart/content fits the new layout proportions

6. Post-edit verification (MANDATORY after ANY modification)

After every edit, reorder, or removal, do ALL of the following before reporting completion:

Import cleanup — scan App.jsx line by line:

  • Every import at the top of App.jsx is actually used in the JSX below. Remove any that aren't (including leftover config, component, or data imports from removed slides)
  • No slide component file imports modules that no longer exist

Storyline review — re-read the full <Deck> in App.jsx and verify:

  • Logical flow: Does each slide follow naturally from the previous one?
  • Progressive disclosure: Concepts introduced before they're referenced?
  • No orphan slides: Every slide belongs to a clear section
  • Title scan test: Read ONLY the slide titles in order — does the story make sense from titles alone?

Report the final slide order with titles to the user so they can confirm.


Sharing as Single HTML

When the user wants to share the deck as a single portable HTML file (for email, IM, or offline viewing):

npm run build:single

This produces dist-single/index.html — a self-contained file with all JS, CSS, and SVG inlined. No server needed; just open in any browser.

How it works: vite.single.config.js uses vite-plugin-singlefile to inline all assets. The logo SVG is imported as a module (not a /public reference) so it gets bundled too.

When to suggest this: If the user says "share", "send to someone", "email the deck", "make it portable", "single file", "offline", or "不需要服务器".

Limitations:

  • File size is ~380KB (ECharts + React bundle) — fine for sharing via email/IM, not ideal for web hosting
  • Custom logos in public/ (e.g., config.logo = '/my-logo.png') are automatically inlined as data URIs by the build plugin — no extra steps needed

Important Rules

  • NEVER fabricate data — all numbers must come from source files
  • When creating a project, generate BOTH CLAUDE.md and AGENTS.md
  • slide-kit.config.js is the single source of truth for runtime config
  • Slide components: src/slides/Slide<PascalCaseName>.jsx
  • Data files: src/data/<kebab-case-name>.js

平台分布

Codex

32.02%
按下载量换算85

Claude

32.37%
按下载量换算85

Cursor

19.14%
按下载量换算51

Gemini CLI

9.71%
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

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