@crashbytes/env-shield

env-shield

TypeScript-first environment variable validation with full type inference. Zero dependencies.

npm: https://www.npmjs.com/package/@crashbytes/env-shield

Why env-shield?

Environment variables are the backbone of modern application configuration — API keys, database URLs, feature flags, ports, and secrets all flow through process.env. Yet most codebases treat them as an afterthought: untyped process.env.WHATEVER calls scattered across files, with no validation, no type safety, and no visibility into what's actually required.

This creates real problems:

• Silent failures in production. A missing DATABASE_URL doesn't crash at startup — it crashes minutes later in a request handler, buried under a confusing stack trace. A typo in STRIPE_SECRT_KEY goes unnoticed until a customer can't check out.
• No type safety. Every process.env access returns string | undefined. You're left casting, parsing, and null-checking the same variables everywhere. Your editor can't help you, and refactoring is guesswork.
• No single source of truth. Which env vars does your app actually need? What format should they be in? Is DEBUG a string or a boolean? Is PORT validated? The answers are scattered across dozens of files, .env.example files that drift out of sync, and tribal knowledge.
• Inconsistent parsing. One file checks process.env.DEBUG === 'true', another checks process.env.DEBUG === '1', a third forgets to check at all. Number parsing is done inline with parseInt or Number() with inconsistent error handling.

What env-shield does differently

env-shield solves all of this with a single schema definition:

import { createEnv, s } from '@crashbytes/env-shield'

const env = createEnv({
  schema: {
    DATABASE_URL: s.url(),
    PORT: s.port().default(3000),
    NODE_ENV: s.enum('development', 'staging', 'production'),
    DEBUG: s.boolean().default(false),
    API_KEY: s.string().min(10),
  },
})

From this one declaration, you get:

• Fail-fast validation at startup. If any variable is missing or malformed, your app crashes immediately with a clear, actionable error message — not five minutes later in an unrelated code path.
• Full TypeScript inference. env.PORT is number, not string | undefined. env.NODE_ENV is 'development' | 'staging' | 'production', not string. No manual type declarations, no casting. Your editor autocompletes everything.
• A single source of truth. One file defines every environment variable your app needs, its type, its constraints, and whether it's required or optional. New team members can read this file and understand the entire configuration surface.
• Consistent parsing and coercion. Booleans accept true/false/1/0/yes/no/on/off. Numbers are parsed and validated. Ports are range-checked. URLs and emails are format-validated. All of this happens once, at startup, in one place.

How it compares

Feature	process.env	envalid	t3-env	env-shield
Type inference	No	Partial	Yes	Yes
Runtime validation	No	Yes	Yes	Yes
Zero dependencies	N/A	No	No (zod)	Yes
Framework agnostic	Yes	Yes	No (Next.js)	Yes
Actively maintained	N/A	Stale	Yes	Yes
Bundle size	0	~15 KB	~60 KB+	~8 KB
Fail-fast errors	No	Yes	Yes	Yes
Built-in validators	No	Limited	Via zod	Yes

env-shield gives you the type safety of t3-env and the validation of envalid, without locking you into a framework or pulling in heavy dependencies. It works everywhere — Next.js, Express, Fastify, AWS Lambda, CLI tools, or any Node.js application.

Install

npm install @crashbytes/env-shield

Quick Start

import { createEnv, s } from '@crashbytes/env-shield'

const env = createEnv({
  schema: {
    DATABASE_URL: s.url(),
    PORT: s.port().default(3000),
    NODE_ENV: s.enum('development', 'staging', 'production'),
    DEBUG: s.boolean().default(false),
    API_KEY: s.string().min(10),
    MAX_RETRIES: s.number().int().min(0).max(10).default(3),
    ADMIN_EMAIL: s.email().optional(),
    FEATURE_FLAGS: s.json<{ darkMode: boolean }>(),
  },
})

// Full type inference - no manual typing needed
env.DATABASE_URL  // string
env.PORT          // number
env.NODE_ENV      // 'development' | 'staging' | 'production'
env.DEBUG         // boolean
env.API_KEY       // string
env.MAX_RETRIES   // number
env.ADMIN_EMAIL   // string | undefined
env.FEATURE_FLAGS // { darkMode: boolean }

Schema Types

Builder	Output Type	Description
s.string()	string	Basic string variable
s.number()	number	Parsed number (integers and floats)
s.boolean()	boolean	Accepts true/false/1/0/yes/no/on/off (case-insensitive)
s.port()	number	Integer between 1 and 65535
s.url()	string	Validated URL format
s.email()	string	Validated email format
s.enum('a', 'b')	'a' | 'b'	Must be one of the specified values
s.json<T>()	T	Parsed JSON string

Modifiers

All schema types support the following chainable modifiers:

// Optionality
s.string().optional()       // value can be missing (type includes undefined)
s.string().default('val')   // fallback value when missing (type excludes undefined)

// String validators
s.string().min(5)           // minimum length
s.string().max(100)         // maximum length
s.string().regex(/^sk-/)    // regex pattern match
s.string().url()            // URL format validation
s.string().email()          // email format validation

// Number validators
s.number().min(0)           // minimum value
s.number().max(1000)        // maximum value
s.number().int()            // must be an integer

// Documentation
s.string().describe('The primary database connection string')

Modifiers can be chained in any order:

s.number().int().min(1).max(100).default(10)
s.string().min(10).regex(/^sk-/).describe('Stripe secret key')

Error Output

When validation fails, env-shield collects all errors and throws a single EnvValidationError with a clear, actionable message:

env-shield: 3 environment variables failed validation:

  ✗ DATABASE_URL: Required variable is missing
  ✗ PORT: Expected a number, got "abc" (received: "abc")
  ✗ NODE_ENV: Must be one of: development, staging, production (received: "test")

No more chasing down one error at a time. Fix everything in a single pass.

Catching errors programmatically

import { createEnv, s, EnvValidationError } from '@crashbytes/env-shield'

try {
  const env = createEnv({ schema: { ... } })
} catch (err) {
  if (err instanceof EnvValidationError) {
    console.error(err.errors) // Array of { key, message, value? }
    process.exit(1)
  }
}

Custom Source

By default, env-shield reads from process.env. You can provide any object as the source:

// Read from a custom object (useful for testing)
const env = createEnv({
  schema: {
    API_KEY: s.string(),
    DEBUG: s.boolean().default(false),
  },
  source: {
    API_KEY: 'test-key-1234567890',
    DEBUG: 'true',
  },
})

Real-World Examples

Express / Fastify API

// src/config.ts
import { createEnv, s } from '@crashbytes/env-shield'

export const env = createEnv({
  schema: {
    PORT: s.port().default(3000),
    HOST: s.string().default('0.0.0.0'),
    NODE_ENV: s.enum('development', 'staging', 'production'),
    DATABASE_URL: s.url(),
    REDIS_URL: s.url().optional(),
    JWT_SECRET: s.string().min(32),
    CORS_ORIGIN: s.url().default('http://localhost:3000'),
    LOG_LEVEL: s.enum('debug', 'info', 'warn', 'error').default('info'),
  },
})

Next.js

// src/env.ts
import { createEnv, s } from '@crashbytes/env-shield'

export const env = createEnv({
  schema: {
    DATABASE_URL: s.url(),
    NEXTAUTH_SECRET: s.string().min(32),
    NEXTAUTH_URL: s.url(),
    STRIPE_SECRET_KEY: s.string().regex(/^sk_/),
    STRIPE_WEBHOOK_SECRET: s.string().regex(/^whsec_/),
    NEXT_PUBLIC_APP_URL: s.url(),
  },
})

AWS Lambda

import { createEnv, s } from '@crashbytes/env-shield'

const env = createEnv({
  schema: {
    TABLE_NAME: s.string(),
    BUCKET_NAME: s.string(),
    SQS_QUEUE_URL: s.url(),
    REGION: s.string().default('us-east-1'),
    LOG_LEVEL: s.enum('debug', 'info', 'warn', 'error').default('info'),
  },
})

License

MIT