Backend Dev

Backend Development

Core Principles

• Stateless services — don't store session state in memory; use Redis or a DB
• Fail fast — validate inputs at the boundary, return errors early
• Idempotency — POST/PUT endpoints that can be called multiple times safely
• Observability — every service needs logs, metrics, and traces

---

API Endpoint Structure

Request Lifecycle

Request → Rate Limiting → Authentication → Authorization → Validation → Business Logic → Response

Never skip this order. Auth before business logic. Validation before processing.

Response Standards

// Success
{ "data": { ... }, "meta": { "page": 1, "total": 100 } }

// Error
{ 
  "error": { 
    "code": "VALIDATION_ERROR",
    "message": "Human-readable description",
    "details": [{ "field": "email", "issue": "Invalid format" }]
  }
}

---

Authentication Patterns

JWT Best Practices

• Short access token TTL (15 min) + long refresh token (7 days)
• Store refresh token in httpOnly cookie (not localStorage)
• Validate signature, expiry, issuer, and audience
• Rotate refresh tokens on use
• Maintain a revocation list or use short-lived tokens only

Session-Based Auth

• Store session ID in httpOnly, Secure, SameSite cookie
• Store session data in Redis with TTL
• Regenerate session ID after login (session fixation prevention)

Authorization

// Check authorization at the service layer, not just the route
async function getOrder(userId: string, orderId: string) {
  const order = await db.orders.findById(orderId)
  if (!order) throw new NotFoundError()
  // Authorization check — user can only see their own orders
  if (order.userId !== userId) throw new ForbiddenError()
  return order
}

---

Background Jobs & Queues

Use queues for:

• Email sending
• Image/file processing
• External API calls that don't need to be synchronous
• Long-running tasks

Job Patterns

Fire-and-forget: Enqueue and return immediately
Scheduled: Run at specific time (cron)
Delayed: Run after X minutes/hours
Retry: Re-attempt on failure with backoff

Always implement:

• Dead letter queue for failed jobs
• Idempotency key to prevent duplicate processing
• Timeout per job
• Max retry count

---

Error Handling

// Structured error hierarchy
class AppError extends Error {
  constructor(
    public message: string,
    public statusCode: number,
    public code: string
  ) { super(message) }
}

class ValidationError extends AppError {
  constructor(details: ValidationDetail[]) {
    super('Validation failed', 400, 'VALIDATION_ERROR')
    this.details = details
  }
}

// Global error handler
app.use((err, req, res, next) => {
  if (err instanceof AppError) {
    return res.status(err.statusCode).json({ error: err })
  }
  // Log unexpected errors
  logger.error('Unexpected error', { err, req })
  res.status(500).json({ error: { code: 'INTERNAL_ERROR' } })
})

---

Middleware Checklist

Every production API needs:

• [ ] Request logging (method, path, status, duration)
• [ ] Rate limiting (per IP, per user, per endpoint)
• [ ] CORS configuration (explicit whitelist)
• [ ] Helmet / security headers
• [ ] Request ID propagation (for distributed tracing)
• [ ] Body size limits
• [ ] Timeout handling

---

Environment Configuration

// Never hardcode configuration — always from environment
const config = {
  db: {
    url: process.env.DATABASE_URL,      // required
    poolSize: parseInt(process.env.DB_POOL_SIZE ?? '10')
  },
  jwt: {
    secret: process.env.JWT_SECRET,     // required
    expiresIn: process.env.JWT_TTL ?? '15m'
  }
}

// Validate required config at startup — fail fast
const required = ['DATABASE_URL', 'JWT_SECRET']
for (const key of required) {
  if (!process.env[key]) throw new Error(`Missing required env: ${key}`)
}

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jamestorrevillas
• Source: jamestorrevillas/dev-skills
• License: MIT

Install and usage instructions live in the source repository linked above.