FE

feature-slicing

Helps structure frontend code into domain-based slices with strict dependency rules.

Install

mkdir -p .claude/skills/feature-slicing && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/19541" && unzip -o skill.zip -d .claude/skills/feature-slicing && rm skill.zip

Installs to .claude/skills/feature-slicing

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Proactively apply when creating new features/components/pages or setting up frontend project structure. Triggers on FSD, feature slicing, Feature-Sliced Design, frontend architecture, layer structure, module boundaries, scalable frontend, slice organization, public API, barrel exports, import rules, Steiger. Use when restructuring React/Next.js/Vue/Remix projects, organizing frontend code, fixing import violations, deciding where code belongs (entity vs feature vs widget vs shared), or migrating legacy codebases. Feature-Sliced Design (FSD) architecture for frontend projects.
582 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Advanced

Key capabilities

  • Enforce layer hierarchy and import rules
  • Organize code by business domain
  • Validate architecture using Steiger
  • Define public APIs for slices
  • Structure frontend projects into app, pages, widgets, features, entities, and shared layers

How it works

It applies the Feature-Sliced Design methodology by restricting imports to strictly lower layers and organizing code into domain-specific slices.

Inputs & outputs

You give it
A request to structure or refactor a frontend project
You get back
A directory structure compliant with FSD layer hierarchy

When to use feature-slicing

  • Structure a new frontend project with FSD
  • Refactor legacy code into business domain slices
  • Enforce layer dependency rules in code
  • Organize frontend components and modules

About this skill

Feature-Sliced Design Architecture

Frontend architecture methodology (spec v2.1) with a strict layer hierarchy and one import rule. FSD organizes code by business domain rather than technical role.

Official Docs: feature-sliced.design | GitHub: feature-sliced


THE IMPORT RULE (Critical)

A module in a slice may only import slices from layers strictly below. Never sideways or upward. This is what keeps slices replaceable and dependencies traceable — every violation you allow becomes an invisible coupling someone else trips over.

app → pages → widgets → features → entities → shared
 ↓      ↓        ↓          ↓          ↓         ✓
 ✓      ✓        ✓          ✓          ✓      (external only)
ViolationExampleFix
Cross-slice (same layer)features/authfeatures/user-profileMove shared code to a lower layer, or compose both in the page/widget above
Upward importentities/userfeatures/authMove the needed code down
Shared importing upshared/entities/Shared depends only on external packages
Cross-entityentities/orderentities/product internalsUse @x notation (entities layer only)

Exception: app/ and shared/ are each "a layer and a slice at the same time" — they divide directly into segments, and their segments may freely import each other.


Layer Hierarchy

LayerPurposeHas SlicesRequired
app/Initialization, routing, providers, global stylesNo (segments only)Yes
pages/Route-based screens — the default home for most codeYesYes
widgets/Large self-sufficient UI blocks reused across pages; may own their data fetching and logicYesNo
features/User interactions reused across pages (login, add-to-cart)YesNo
entities/Business domain models reused across pages (user, product)YesNo
shared/Reused infrastructure: UI kit, API client, route constants, utilities. No knowledge of domain slices — but it may be app-aware (your backend's client, your route paths)No (segments only)Yes

The processes/ layer is deprecated — move its contents to features/ and app/.

Minimal setup: app/, pages/, shared/. A small app with 5 routes does not need entities, features, or widgets — adding them "for completeness" spreads single-use code across the tree and negates FSD's cohesion benefit. Add a layer only when something is genuinely reused by 2+ pages.


Pages First (v2.1 default)

When unsure where code goes, put it in the page slice that uses it — including forms, data logic, and large UI blocks. Extract downward only when a second page needs it. Why: page-local code is found instantly by newcomers; premature entities/features force jumping through several folders to change one user flow.

Where does this code go?
├─ Used by exactly ONE page              → that page's slice (even logic/forms)
├─ App-wide config, providers, routing   → app/
├─ Domain-agnostic or infra code         → shared/
├─ Reused across pages:
│  ├─ Large UI block with own logic      → widgets/
│  ├─ User action (verb)                 → features/
│  └─ Domain model/data (noun)           → entities/

"Feature or Entity?"

Entity (noun)Feature (verb)
user — user data modelauth — login/logout actions
product — product infoadd-to-cart — adding to cart
comment — comment datawrite-comment — creating comments

Entities represent THINGS with identity. Features represent ACTIONS with side effects. Not everything is a feature — an interaction used on one page stays in that page.

"Which segment?"

Segments divide a slice by technical purpose:

├─ ui/      → components, styles, formatters
├─ api/     → backend calls, DTOs, mappers
├─ model/   → types, schemas, stores, business logic
├─ lib/     → slice-internal utilities
└─ config/  → feature flags, constants

Name segments by purpose, not essence: api/, model/, lib/ — never hooks/, components/, types/, utils/. Purpose names tell a reader what the code is for; essence names just restate the file extension.


Directory Structure

src/
├── app/                    # Layer + slice: segments only
│   ├── providers/          # React context, QueryClient, theme
│   ├── routes/             # Router configuration
│   └── styles/             # Global CSS, theme tokens
├── pages/
│   └── {page-name}/
│       ├── ui/             # Page component + page-local blocks
│       ├── api/            # Loaders, server actions
│       ├── model/          # Page-specific state
│       └── index.ts        # Public API
├── widgets/
│   └── {widget-name}/
│       ├── ui/
│       ├── api/            # Widgets may fetch their own data (v2.1)
│       └── index.ts
├── features/
│   └── {feature-name}/
│       ├── ui/
│       ├── api/
│       ├── model/
│       └── index.ts
├── entities/
│   └── {entity-name}/
│       ├── ui/             # Entity UI (Card, Avatar)
│       ├── api/            # CRUD operations
│       ├── model/          # Types, mappers, validation
│       └── index.ts
└── shared/                 # Layer + slice: segments only, no root index
    ├── ui/                 # Design system (index per component)
    ├── api/                # API client, interceptors
    ├── lib/                # Utilities (dates, validation)
    ├── config/             # Environment, constants
    ├── routes/             # Route path constants
    └── i18n/               # Translations

Public API Pattern

Every slice exposes ONE public API via index.ts. External code imports only from it — the index is the contract that lets a slice refactor its internals freely.

// entities/user/index.ts
export { UserCard } from './ui/UserCard';
export { getUser, updateUser } from './api/userApi';
export type { User, UserRole } from './model/types';
// ✅ Correct
import { UserCard, type User } from '@/entities/user';

// ❌ Wrong — reaches into internals
import { UserCard } from '@/entities/user/ui/UserCard';

Three rules that prevent the common failure modes:

  1. Explicit named exports, no wildcards. export * from './ui' hides the contract, leaks internals, and harms tree-shaking.
  2. No segment-level index files on sliced layers. If features/comments/index.ts exists, don't also create features/comments/ui/index.ts — extra barrels slow bundlers and invite circular imports. Inside a slice, use relative imports; never import your own index.ts.
  3. On shared/, the public API lives per segment (shared/api/index.ts), and for shared/ui/shared/lib, per component (shared/ui/button/index.ts) — one giant barrel bundles a syntax highlighter into every page.

Details and edge cases: references/PUBLIC-API.md.


Cross-Entity References (@x Notation)

Entities sometimes genuinely contain each other (an order contains products). Instead of a hidden cross-import, make it explicit with @x — used only on the entities layer:

entities/
├── product/
│   ├── @x/
│   │   └── order.ts    # exports intended specifically for the order entity
│   └── index.ts
└── order/
    └── model/types.ts  # imports from product/@x/order
// entities/product/@x/order.ts
export type { ProductId } from '../model/types';

// entities/order/model/types.ts
import type { ProductId } from '@/entities/product/@x/order';

Keep @x files tiny (usually type re-exports). If two entities cross-reference extensively, merge them.


Anti-Patterns

Anti-PatternProblemFix
Business logic in shared/shared/lib/calculateDiscount.ts knows domain rulesMove to the entity/feature/page that owns it; shared holds infra only
Cross-slice import (same layer)Invisible coupling between siblingsExtract down, use @x (entities), or compose in the layer above
Over-slicing a small app6 layers, 20 slices, every slice used oncePages-first: keep code in page slices; layers earn their existence via reuse
Everything is a featurefeatures/close-modalOnly reused, business-meaningful interactions
Single-use widgetsWidget used by one pageKeep in that page's ui/
Generic segmentscomponents/, hooks/, utils/Purpose names: ui/, model/, lib/
Wildcard exportsexport * from './ui'Explicit named exports
Bypassing public APIDeep imports into a sliceImport from the slice index.ts

Tooling

Verify architecture automatically with Steiger, the official FSD linter:

npm i -D steiger @feature-sliced/steiger-plugin
npx steiger ./src            # add --watch during refactors

It catches cross-imports, missing public APIs, and single-use slices (fsd/insignificant-slice). ESLint alternative: @feature-sliced/eslint-config.

Path alias (required for @/entities/user-style imports):

{
  "compilerOptions": {
    "baseUrl": ".",
    "paths": { "@/*": ["./src/*"] }
  }
}

Reference Documentation

Read thisWhen
references/LAYERS.mdDeciding which layer code belongs to; full layer specs and flowchart
references/PUBLIC-API.mdWriting index files, fixing circular imports, tree-shaking, @x details
references/IMPLEMENTATION.mdWriting actual code: complete entity/feature/widget/page examples with React Query, Zustand, Zod
references/NEXTJS.mdAny Next.js project — the app/pages folder conflict, _app/_pages renaming, server/client public APIs
[references/M

Content truncated.

When not to use it

  • Adding layers for completeness in small apps
  • Using essence-based segment names like components or utils

Prerequisites

Steiger linterPath alias configuration

Limitations

  • Modules cannot import sideways or upward
  • Shared layer depends only on external packages

How it compares

Unlike generic folder structures, it enforces a strict hierarchy where modules can only import from layers below them to prevent coupling.

Compared to similar skills

feature-slicing side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
feature-slicing (this skill)020dReviewAdvanced
nuxt04moNo flagsIntermediate
ai-model-web11moReviewIntermediate
moai-domain-frontend12moNo flagsAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

nuxt

onmax

Always-on Nuxt disambiguation layer for this project. Use it to choose the right Nuxt pack first, then switch to Vue or module guidance only when Nuxt no longer owns the abstraction.

00

ai-model-web

TencentCloudBase

Use this skill when developing browser/Web applications (React/Vue/Angular, static websites, SPAs) that need AI capabilities. Features text generation (generateText) and streaming (streamText) via @cloudbase/js-sdk. Built-in models include Hunyuan (hunyuan-2.0-instruct-20251111 recommended) and DeepSeek (deepseek-v3.2 recommended). NOT for Node.js backend (use ai-model-nodejs), WeChat Mini Program (use ai-model-wechat), or image generation (Node SDK only).

13

moai-domain-frontend

modu-ai

Frontend development specialist covering React 19, Next.js 16, Vue 3.5, and modern UI/UX patterns with component architecture. Use when building web UIs, implementing components, optimizing frontend performance, or integrating state management.

13

perf-web-optimization

tech-leads-club

Optimize web performance: Core Web Vitals (LCP, CLS, INP), bundle size, images, caching. Use when site is slow, optimizing for Lighthouse scores, reducing bundle size, fixing layout shifts, or improving Time to Interactive. Triggers on: web performance, Core Web Vitals, LCP, CLS, INP, FID, bundle size, page speed, slow site.

10

rdc-setup

reactive

Install and set up @data-client/react or @data-client/vue in a project. Detects project type (NextJS, Expo, React Native, Vue, plain React) and protocol (REST, GraphQL, custom), then hands off to protocol-specific setup skills.

10

astro

Anhvu1107

ALWAYS use this when the request matches Astro: Build content-focused websites with Astro — zero JS by default, islands architecture, multi-framework components, and Markdown/MDX support.

00

Search skills

Search the agent skills registry