前端审查
- 作者仓库星标 3
- 作者更新于 实时读取
- 作者仓库 dotnix
- 领域
- 设计与多媒体
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @yutakobayashidev · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: functional-cohesion-components
description: Guide frontend component design and code review using functional cohesion. Use when designing, r…
category: 设计与多媒体
runtime: 无特殊运行时
---
# functional-cohesion-components 输出预览
## PART A: 任务判断
- 适用问题:视觉内容、演示材料、信息图或设计交付。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Purpose / Core Rule / Review Workflow”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于视觉内容、演示材料、信息图或设计交付,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Purpose / Core Rule / Review Workflow”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件。
先用一个小任务确认它会围绕“Purpose / Core Rule / Review Workflow”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: functional-cohesion-components
description: Guide frontend component design and code review using functional cohesion. Use when designing, r…
category: 设计与多媒体
source: yutakobayashidev/dotnix
---
# functional-cohesion-components
## 什么时候使用
- 用于审阅代码、文档或方案并给出可执行反馈 适合处理界面、视觉、封面、信息图或演示材料交付,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代的步骤;通常不需要额外 A…
- 面向视觉内容、演示材料、信息图或设计交付,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Purpose / Core Rule / Review Workflow」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "functional-cohesion-components" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Purpose / Core Rule / Review Workflow
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Functional Cohesion Components
Purpose
Use functional cohesion to keep frontend components aligned with product responsibilities. Prefer components that represent one meaningful feature, role, or workflow over components that switch many responsibilities with flags.
Use this skill to answer: "Should this UI be one component with conditions, split into multiple components, or partially commonized?"
Core Rule
Prefer functional cohesion:
- Split by responsibility when UI represents different roles, workflows, permissions, states, or user intentions.
- Commonize only a meaningful unit, not a visually similar fragment.
- Keep unavoidable branching near the boundary that decides what feature is being rendered.
- Avoid logical cohesion: one component that groups similar-looking behavior and switches responsibilities with
role,type, or many boolean flags.
Do not apply this mechanically. If a split creates components with almost no responsibility, keep the component together and use a small feature-named extension point.
Review Workflow
Identify the feature boundary. Ask what product responsibility the component represents: buyer view, seller view, admin operation, notification kind, file row, directory row, create page, edit page, etc.
List the differences. Check displayed data, actions, navigation target, permissions, validation, empty/error states, mobile behavior, and future variation risk.
Decide where branching belongs. Prefer route/page boundaries first. If data type decides the branch, put the branch in a parent, list mapper, or small dispatcher component. Keep child components focused.
Evaluate commonization. Commonize only when the unit is also meaningfully shared in requirements, Figma grouping/componentization, API schema shape, and local code conventions.
Check for over-splitting. If a split only wraps one extra line around a large shared component, use a small feature-named prop or slot instead.
State the tradeoff. In reviews, say which responsibility is being isolated, what future conditions are avoided, and what duplication is intentionally accepted.
Patterns
Route-Separable Screens
When routing already separates roles or page purposes, let the route call different page components.
Good examples:
BuyerProductPageandSellerProductPagefrom role-specific routes.CreateProductPageandEditProductPageas separate pages.- A shared
ProductFormonly when schema and field responsibility are truly the same.
For create/edit forms, keep title, default values, and submit handler in the page. Let the form render fields and invoke passed handlers.
Avoid hiding route-level responsibilities behind a role prop:
export const ProductDetail = ({
product,
role,
}: {
product: Product;
role: 'buyer' | 'seller';
}) => (
<div>
<h2>{product.name}</h2>
<p>{product.description}</p>
{role === 'buyer' && <button>Add to cart</button>}
{role === 'seller' && <button>Edit</button>}
</div>
);
Prefer role-specific page or feature components when the role has its own responsibility:
export const BuyerProductDetail = ({ product }: { product: Product }) => (
<div>
<ProductBasicInfo product={product} />
<button>Add to cart</button>
</div>
);
export const SellerProductDetail = ({ product }: { product: Product }) => (
<div>
<ProductBasicInfo product={product} />
<button>Edit</button>
</div>
);
const ProductBasicInfo = ({ product }: { product: Product }) => (
<section>
<h2>{product.name}</h2>
<p>{product.description}</p>
</section>
);
For create/edit, share the form only after moving page-specific responsibility upward:
export const CreateProductPage = () => (
<ProductForm
title="Add New Product"
defaultValues={{ name: '', price: '', description: '' }}
submitLabel="Create"
onSubmit={createProduct}
/>
);
export const EditProductPage = ({ product }: { product: Product }) => (
<ProductForm
title="Edit Product"
defaultValues={product}
submitLabel="Edit"
onSubmit={(values) => updateProduct(product.id, values)}
/>
);
Data-Type Branching
When API data determines the UI type, a branch is unavoidable. Do not scatter the same branch across icon, click handler, menu items, and layout.
Prefer:
- A parent dispatcher such as
FileSystemRow. - A
list.map()branch that choosesDirectoryRoworFileRow. - Exhaustive matching for discriminated unions.
Use TypeScript discriminated unions or ts-pattern with .exhaustive() when the API schema defines variants such as notification kinds. This makes new backend variants fail type-checking until the UI handles them.
Avoid repeating the same discriminator throughout one row:
const StorageItemRow = ({ item }: { item: StorageItem }) => (
<tr onClick={() => (item.type === 'folder' ? openFolder(item.id) : previewFile(item.id))}>
<td>
{item.type === 'folder' ? <FolderIcon /> : <FileIcon />} {item.name}
</td>
<td>
{item.type === 'file' && <button onClick={() => downloadFile(item.id)}>Download</button>}
{item.type === 'file' && <button onClick={() => deleteFile(item.id)}>Delete</button>}
</td>
</tr>
);
Prefer one branch at the boundary, then focused rows:
const StorageItemRow = ({ item }: { item: StorageItem }) => {
switch (item.type) {
case 'folder':
return <FolderRow folder={item} />;
case 'file':
return <FileRow file={item} />;
}
};
For notifications, let exhaustive matching express that each notification kind is its own feature:
const NotificationItem = ({ notification }: { notification: Notification }) =>
match(notification)
.with({ type: 'OrderCreated' }, ({ orderId, createdAt }) => (
<OrderCreatedNotification orderId={orderId} createdAt={createdAt} />
))
.with({ type: 'ProductLiked' }, ({ productId, createdAt }) => (
<ProductLikedNotification productId={productId} createdAt={createdAt} />
))
.with({ type: 'MessageReceived' }, ({ chatId, createdAt }) => (
<MessageReceivedNotification chatId={chatId} createdAt={createdAt} />
))
.exhaustive();
Commonization Pressure
When two areas look similar, verify they are functionally related before sharing code.
Use these signals:
- Requirements describe the same concept.
- Figma groups or components the same unit.
- API schema uses the same object shape for the same domain concept.
- Existing codebase convention splits at the same grain.
Be skeptical when a buyer-facing history screen resembles an admin order-management screen, or when User and SellerUser look similar but encode different concepts.
Small Differences
When differences are tiny, splitting by role can create thin and unbalanced components.
Accept a small feature-named prop when it is local and limited:
<ProductDetail product={product} showEditButton />
Prefer names like showEditButton, showStockWarning, or extraProductInfo over role="seller" or type="admin". One or two such options can be reasonable; many options indicate the component wants to split.
Use ReactNode or children slots sparingly. Name the slot after the feature, not the layout position, because the child component cannot otherwise reveal what is injected.
Use a feature-named prop for a tiny local variation:
export const ProductDetail = ({
product,
showEditButton = false,
}: {
product: Product;
showEditButton?: boolean;
}) => (
<div>
<h2>{product.name}</h2>
<p>{product.description}</p>
{showEditButton && <button>Edit</button>}
</div>
);
Use a named slot only when the parent owns the extra feature and the base component still has a clear responsibility:
export const ProductDetail = ({
product,
extraProductInfo,
}: {
product: Product;
extraProductInfo?: React.ReactNode;
}) => (
<div>
<h2>{product.name}</h2>
<p>{product.description}</p>
{extraProductInfo}
</div>
);
Warning Signs
Treat these as review findings:
- The same
roleortypecheck appears in several places inside one component. - Boolean flags describe roles or modes instead of concrete features.
- A "shared" component needs comments explaining which role owns which lines.
- Adding a new role requires auditing unrelated branches.
- UI similarity is the only reason for commonization.
- A child component mixes navigation, permissions, and rendering for multiple domain concepts.
- Branching exists at both parent and child levels for the same discriminator.
Review Output
When reviewing code, structure the answer as:
- Current cohesion: functional, logical, or mixed.
- Main risk: the concrete branch or shared unit likely to grow.
- Recommended boundary: route/page, dispatcher, row/component, form, or feature prop.
- Duplication decision: what to duplicate intentionally and what to share.
- Verification: type exhaustiveness, affected role flows, and tests or story cases to cover.
Keep recommendations simple. Prefer the smallest change that moves responsibility boundaries closer to product meaning.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核