AGENTS.md

Agent guidelines

Haxe coding style

Prefer modern Haxe syntax where it fits:

• Use final for readonly fields and locals that are not reassigned.
• Prefer arrow function expressions (() -> …) for short callbacks and lambdas.
• Prefer named-argument function types ((name:Type) -> Ret) over positional-only forms when declaring function types.
• Use type-safe values from why-unit library where applicable

Functional style

Prefer a functional programming paradigm:

• Prefer pure functions: same inputs → same outputs, no hidden mutation or I/O.
• Keep side effects (I/O, mutable state, process/env interaction) clearly isolated at the edges — e.g. command handlers, host adapters — not scattered through core logic.
• Prefer transforming data with expressions (map, filter, immutable updates) over imperative mutation when practical.

Explicit member types

All member functions must declare argument and return types explicitly.

Exception: lambdas / closures may omit types when they are inferred from context.

// ✅ GOOD
public function measure(fn:() -> Void, ?opts:MeasureOptions):MeasureResult { … }

final sink = (v:Any) -> { /* … */ };
items.map(x -> x * 2);

// ❌ BAD — untyped member
public function measure(fn, ?opts) { … }