missing-throw.md

llm-core/missing-throw

📝 Disallow standalone new Error expressions that likely forgot the throw keyword.

💼 This rule is enabled in the following configs: 🌐 all, 🏆 best-practices, ✅ recommended.

🔧 This rule is automatically fixable by the --fix CLI option.

Rule details

This rule flags standalone new Error(...) expressions that create an Error object without throwing, returning, assigning, or otherwise using it. new Error(...) by itself does not stop execution. The autofix inserts the missing throw keyword before the Error construction.

Flagged patterns:

function requireUser(user) {
  if (!user) {
    new Error("User is required");
  }
}

function parse(value) {
  if (typeof value !== "string") {
    new TypeError("Expected string");
  }
}

function collectErrors(errors) {
  new AggregateError(errors, "Multiple failures");
}

Correct patterns:

function requireUser(user) {
  if (!user) {
    throw new Error("User is required");
  }
}

function createError() {
  return new Error("User is required");
}

const error = new Error("User is required");

Scope and limitations

This rule uses AST-only analysis. It flags standalone construction of built-in Error constructors:

• Error
• AggregateError
• EvalError
• RangeError
• ReferenceError
• SyntaxError
• TypeError
• URIError

It intentionally does not flag arbitrary custom classes such as new ValidationError(...) because some codebases construct custom error objects as values and the rule cannot prove they are error subclasses without type information.