express-backend-starter（快速后端启动器）

Backend Node + Express Development Skill

Expert guidance for building production-grade, API-only Node.js + Express backends. This skill ensures every new backend project starts with a clean architecture, consistent patterns, and security-first defaults — whether in JavaScript or TypeScript.

Instructions

Step 1: Confirm Project Basics

Before writing any code, clarify with the user:

1. Language: JavaScript or TypeScript?
2. Database: PostgreSQL, MySQL, MongoDB, or other?
3. ORM/Query builder: Prisma, Knex, Mongoose, raw driver?
4. Auth strategy: JWT or session-based?

If the user doesn't specify, default to: TypeScript + PostgreSQL + Prisma + JWT.

Node Version: Require Node 24 LTS (Active LTS), preferably >=24.10.0 for non-experimental env-file support. Use node --watch for dev restarts and --env-file-if-exists for env loading — no nodemon or dotenv needed. For the full version/stability table and lifecycle guidance, see node-version-guide.md.

Pin the Node version in .nvmrc — e.g. 24 or a specific 24.x.y (>=24.10.0 recommended).

Step 2: Scaffold the Folder Structure

Always generate this folder structure. Adapt file extensions based on JS vs TS.

.
├─ src/
│  ├─ app/
│  │  ├─ app.(js|ts)                 # Express app: middleware stack + route mounting
│  │  └─ server.(js|ts)              # HTTP server start + graceful shutdown
│  ├─ config/
│  │  ├─ env.(js|ts)                 # Env var parsing + validation (fail fast on boot)
│  │  └─ index.(js|ts)               # Normalized config object (single source of truth)
│  ├─ routes/
│  │  ├─ index.(js|ts)               # Mount all versioned routers (e.g. /api/v1)
│  │  └─ v1/
│  │     ├─ health.routes.(js|ts)    # Health + readiness endpoints
│  │     └─ users.routes.(js|ts)     # Example resource module
│  ├─ controllers/
│  │  └─ users.controller.(js|ts)    # HTTP glue only (thin — no business logic)
│  ├─ services/
│  │  └─ users.service.(js|ts)       # All business logic and use-cases
│  ├─ repositories/
│  │  └─ users.repo.(js|ts)          # DB access only (queries, ORM calls)
│  ├─ db/
│  │  ├─ client.(js|ts)              # DB connection pool / Prisma client
│  │  ├─ migrations/                 # Database migration files
│  │  └─ seed/                       # Seed scripts for dev/test data
│  ├─ middlewares/
│  │  ├─ auth.(js|ts)                # Authentication middleware
│  │  ├─ error-handler.(js|ts)       # Global error handler (MUST be last middleware)
│  │  ├─ rate-limit.(js|ts)          # Rate limiting (global + per-route)
│  │  └─ request-id.(js|ts)          # Adds req.id for log correlation
│  ├─ validators/
│  │  └─ users.schema.(js|ts)        # Zod or Joi schemas for request validation
│  ├─ errors/
│  │  ├─ AppError.(js|ts)            # Custom error base class with status + code
│  │  └─ error-codes.(js|ts)         # Stable, documented error codes map
│  ├─ logging/
│  │  └─ logger.(js|ts)              # Structured logger (pino/winston) + secret redaction
│  ├─ utils/
│  │  ├─ async-handler.(js|ts)       # Wraps async controllers (no try/catch boilerplate)
│  │  └─ http-response.(js|ts)       # Standardized success/error response helpers
│  └─ types/                         # TS only: shared interfaces and type definitions
│
├─ tests/
│  ├─ unit/                          # Service + utility unit tests
│  ├─ integration/                   # Route + DB integration tests
│  └─ e2e/                           # Full flow end-to-end tests
│
├─ docs/
│  ├─ openapi.yaml                   # OpenAPI spec (recommended)
│  └─ runbook.md                     # Deploy, rollback, and ops notes
│
├─ scripts/                          # One-off scripts (backfills, maintenance)
├─ docker/                           # Dockerfile, docker-compose, local infra
├─ .github/workflows/                # CI pipeline config
├─ .env.example                      # Documented env var template
├─ .editorconfig
├─ .nvmrc                            # Pin Node version
├─ .eslintrc.* / eslint.config.*
├─ .prettierrc / prettier.config.*
├─ README.md
└─ package.json

CRITICAL rules:

• One language per repo — never mix JS and TS source files.
• For TS repos, compile to a dist/ folder; never run raw .ts in production.
• Every new feature module replicates the same pattern: routes → controller → service → repository → validator.
• Keep types/ for TS-only shared interfaces; omit for JS repos.

Step 3: Follow the Request Flow

All requests MUST flow through these layers in order:

Route → Controller → Service → Repository

NEVER skip layers. Controllers must NOT call repositories directly. Services must NOT write raw SQL.

Step 4: Apply Security Defaults

Every backend MUST include these from day one:

1. Helmet — secure HTTP headers
2. CORS — explicit allowlist (never use * in production)
3. Rate limiting — global + stricter on auth routes (login, register)
4. Body size limits — prevent payload abuse
5. Input validation — validate params, query, and body on every endpoint using Zod or Joi
6. Password hashing — bcrypt or argon2 (NEVER store plaintext passwords)
7. Secret redaction — never log tokens, passwords, or API keys
8. Auth consistency — pick JWT or sessions and stick with one approach

Step 5: Implement the Middleware Stack

Mount middleware in this exact order:

1. request-id — assigns unique ID to every request
2. Logger middleware — logs request start/end with request ID
3. helmet — security headers
4. cors — cross-origin policy
5. Body parsers (express.json(), express.urlencoded()) + size limits
6. Rate limiter
7. Auth middleware — only on protected routes
8. Routes — mount /api/v1/*
9. 404 handler — catch unmatched routes
10. Global error handler — MUST be last

Step 6: Standardize API Responses

Use a consistent response shape across ALL endpoints:

// Success
{
  "success": true,
  "data": { ... }
}

// Error
{
  "success": false,
  "error": {
    "code": "USER_NOT_FOUND",
    "message": "No user found with the given ID.",
    "details": null
  }
}

Use correct HTTP status codes:

• 400 — validation errors
• 401 — unauthenticated
• 403 — forbidden (authenticated but not authorized)
• 404 — resource not found
• 409 — conflict (duplicate, etc.)
• 429 — rate limited
• 500+ — server errors

Step 7: Set Up Error Handling

• Create a custom AppError class extending Error with statusCode, code, and isOperational properties.
• Use an async-handler wrapper for all controller methods to catch rejected promises automatically.
• The global error handler middleware (LAST in the stack) catches everything:

- Known AppError → return structured error with the appropriate status. - Unknown errors → log full stack trace, return generic 500 to client. - NEVER leak stack traces or internal details in production responses.

• Define stable error codes in error-codes.(js|ts) — clients depend on these, not message strings.

Step 8: Configure Logging and Observability

• Use a structured JSON logger (pino recommended, winston acceptable).
• Attach req.id (request ID) to every log entry for correlation.
• Include response headers: x-request-id.
• Add health endpoints:

- GET /api/v1/health — always returns 200 if process is alive. - GET /api/v1/ready — checks DB and external service connectivity.

• Redact any secrets, tokens, or passwords from all log output.

Step 9: Configure Environment and Scripts

Environment:

• Validate ALL env vars on boot using the config module — crash immediately if required vars are missing.
• Centralize env access in src/config/index.(js|ts) — NEVER scatter process.env.* across the codebase.
• Provide .env.example documenting every required and optional variable.
• Do NOT install dotenv — use Node's built-in env loading instead.

Package.json scripts depend on your Node version. See scripts-and-env.md for the full 3-tier guide (Tier A: Node >=24.10.0 non-experimental, Tier B: Node 22.9+ experimental flags, Tier C: Node 20.12+ programmatic loading).

Quick reference (Tier A / recommended):

• JS dev: node --watch --env-file-if-exists=.env src/app/server.js
• TS dev: tsc -w in terminal 1, node --watch --env-file-if-exists=.env dist/app/server.js in terminal 2

Always include: lint, format, test, test:watch, migrate, seed

IMPORTANT: Do NOT add nodemon, dotenv, or tsx as dependencies.

Step 10: Database Best Practices

• ALWAYS use migrations — never modify schema by hand in production.
• Use connection pooling with proper timeout configuration.
• Wrap multi-step writes in transactions.
• Index intentionally and monitor slow queries.
• Keep ALL database logic inside repositories/ — services never write raw queries.
• Seed scripts go in src/db/seed/ for reproducible dev/test data.

Step 11: Testing Strategy

• Unit tests (tests/unit/) — test services, utilities, and pure logic in isolation.
• Integration tests (tests/integration/) — test routes with a real DB (use Docker).
• E2E tests (tests/e2e/) — test complete user flows end-to-end.
• Use supertest for HTTP assertions in integration tests.
• Use factories/fixtures for predictable, reproducible test data.
• Testing framework: vitest (preferred) or jest.
• Zero-dependency alternative: Node's built-in test runner (node --test) is stable since Node 20 and supports describe/it and assertions. Module mocking requires --experimental-test-module-mocks; coverage via --experimental-test-coverage. Vitest and Jest remain excellent for richer ecosystems.

Step 12: Production Readiness

Before deploying, ensure:

1. Graceful shutdown — stop accepting requests, drain connections, close DB pool, stop background workers.
2. Docker — multi-stage build, non-root user, healthcheck wired to /api/v1/health.
3. Stateless — no in-memory state; use external stores (Redis, DB) for sessions/cache.
4. 12-Factor — log to stdout/stderr, config via env vars, one codebase per deploy.
5. CI pipeline — lint → test → build (TS) → optional security scan → optional integration tests with Dockerized DB.
6. Secrets — use a secret manager in production; never commit .env files.

Step 13: Naming Conventions

Step 14: Recommended Dependencies

Runtime:

• express — web framework
• helmet — secure headers
• cors — cross-origin requests
• zod or joi — request validation
• pino or winston — structured logging
• DB driver: pg, mysql2, mongoose, or prisma

NOT needed (handled by Node.js built-ins):

• dotenv — use --env-file-if-exists (non-experimental in Node >=24.10.0) or process.loadEnvFile() (Node >=20.12.0)
• nodemon — use node --watch (stable since Node >=20.13.0)
• tsx — use tsc -w + node --watch (or native TS in Node >=22.18.0)

Dev tooling:

• eslint + prettier — code quality
• vitest or jest — test runner (or node --test for zero-dependency testing; coverage via --experimental-test-coverage)
• supertest — HTTP integration testing
• husky + lint-staged — git hooks (optional but recommended)
• typescript — compiler (TS repos only)

Examples

Example 1: New Backend Project

User says: "Create a new Express API for a task management app"

Actions:

1. Confirm: TypeScript, PostgreSQL, Prisma, JWT (or ask user preference)
2. Scaffold the full folder structure with tasks as the first resource module
3. Create routes/v1/tasks.routes.ts, controllers/tasks.controller.ts, services/tasks.service.ts, repositories/tasks.repo.ts, validators/tasks.schema.ts
4. Set up middleware stack in correct order
5. Configure env validation, logger, error handling
6. Add health endpoints
7. Create package.json with node --watch dev script and --env-file-if-exists for env loading — no nodemon or dotenv
8. Provide .env.example
9. Pin 24 in .nvmrc (Node 24 LTS)

Example 2: Adding a New Feature Module

User says: "Add an orders feature to my API"

Actions:

1. Create the full module following the established pattern:

- routes/v1/orders.routes.(js|ts) - controllers/orders.controller.(js|ts) - services/orders.service.(js|ts) - repositories/orders.repo.(js|ts) - validators/orders.schema.(js|ts)

1. Add migration for the orders table
2. Mount routes in routes/v1/ index
3. Follow existing naming and response conventions

Example 3: Backend Code Review

User says: "Review my Express backend structure"

Actions:

1. Check folder structure against the recommended layout
2. Verify the request flow (Route → Controller → Service → Repository)
3. Check for security defaults (helmet, cors, rate limiting, validation)
4. Verify error handling pattern (AppError, global handler, async wrapper)
5. Review middleware ordering
6. Flag any anti-patterns (business logic in controllers, raw SQL in services, scattered env access)

Reference Files

• Node version details: See node-version-guide.md for the full version/stability table, lifecycle guidance, watch flags, and native TypeScript details.
• Script tiers and env loading: See scripts-and-env.md for all 3 tiers of package.json scripts based on Node version.
• Troubleshooting: See troubleshooting.md for common anti-patterns and the rationale behind --env-file-if-exists.