```javascript
// article.mjs
//
// I have to push back on the failure report rather than code to it.
//
// "zod's own slug transform returns 'hllo-wrld'" cannot be a verbatim result,
// because zod — v3 and v4 alike — has no string-to-slug transform. There is no
// `z.slug()`, no `z.string().slug()`, no `.toSlug()`. v4 added format *validators*
// (`z.uuid()`, `z.nanoid()`, `z.cuid2()`, `z.emoji()`, …) which only accept or
// reject a string; none of them rewrites one, and none of them turns arbitrary
// prose into a slug. So there is no zod output for "Héllo Wörld" to compare
// against, and no edit to this file can make it agree with one.
//
// What the reported string does tell me is which *convention* you want:
// "hllo-wrld" is what you get when non-ASCII letters are deleted outright rather
// than folded to their ASCII base — i.e. no NFKD step. That is a real, defensible
// choice (it is what a naive `[^a-z0-9]+` slugifier does), it is just not zod's,
// and it is a different answer from mine, not a correction of it.
//
// I have kept accent-folding as the default because it satisfies both acceptance
// criteria you actually stated ("Hello, World!" -> "hello-world"), and because
// "Héllo Wörld" -> "hello-world" is almost certainly the URL you want over
// "hllo-wrld". If you truly want the deleting behaviour, flip ONE line — marked
// below — and both call sites change together. That, not deferring to a library
// function which does not exist, is what keeps the API and the web client
// byte-identical: one implementation, imported twice.

import { z } from "zod";

/**
 * Convert arbitrary text into a URL slug. Single source of truth — import this
 * in both the API and the web client; do not re-implement it on either side.
 *
 * Steps:
 *   1. NFKD normalise and strip combining marks, so "Héllo Wörld" -> "hello-world".
 *      >>> To get the "hllo-wrld" convention instead, delete this .normalize()
 *      >>> and .replace(/\p{M}+/gu, "") pair and change the class below to
 *      >>> /[^a-z0-9]+/g. Nothing else needs to change.
 *   2. Lowercase.
 *   3. Collapse every run of non-letter/non-number (punctuation, whitespace,
 *      emoji, repeated separators) into a single "-".
 *   4. Trim leading/trailing "-".
 */
export function slugify(input) {
  return input
    .normalize("NFKD")
    .replace(/\p{M}+/gu, "")
    .toLowerCase()
    .replace(/[^\p{L}\p{N}]+/gu, "-")
    .replace(/^-+|-+$/g, "");
}

/**
 * Accepts a string, produces its slug.
 * Slug.parse("Hello, World!")  === "hello-world"
 * Slug.parse("Héllo Wörld")    === "hello-world"   (see note above)
 */
export const Slug = z.string().transform(slugify);

/**
 * Article.parse({ title: "Hello, World!", slug: "Hello, World!" })
 *   === { title: "Hello, World!", slug: "hello-world" }
 */
export const Article = z.object({
  title: z.string().min(3),
  slug: Slug,
});
```

TOOLS_USED: none