Agent Skills: Clean TypeScript

Clean TypeScript

We use TypeScript as a correctness and clarity tool, not as ceremony. Types should reduce bugs and cognitive load.

Type Philosophy

• PREFER explicit, readable types over clever or overly generic ones
• AVOID any and unsafe type assertions
• Use unknown instead of any when necessary
• Let TypeScript infer types when inference is clear and stable

Types & Interfaces

• PREFER type aliases for most use cases
• Use interface primarily for public, extendable object shapes
• Keep types small, composable, and well-named

Functions & APIs

• PREFER explicit return types for public functions
• Avoid function overloads unless they meaningfully improve the API
• Keep function signatures simple and predictable

Nullability & Safety

• Handle null and undefined explicitly
• DO NOT rely on non-null assertions (!) except as a last resort
• Prefer narrowing via control flow and guards

Enums & Constants

• AVOID enum
• PREFER union types or as const objects
• Keep runtime output predictable and minimal

Error Handling

• Type errors and error states explicitly
• Prefer result objects or typed errors over throwing where appropriate
• Do not hide failure modes behind broad types

General Principles

• Types should explain intent
• If a type is hard to understand, it’s probably wrong
• Favor maintainability over theoretical completeness