fix(ui): leave no angle bracket in a sanitized origin; judge markup on the raw one (#16)

* fix(ui): leave no angle bracket in a sanitized origin; judge markup on the raw one

sanitizeOrigin removed tags in a single pass, so `<scr<script>ipt>`
came out as `<script` (CodeQL js/incomplete-multi-character-sanitization
#6, #7). The page renders the result as React text, so this was not
exploitable here, but the helper's name promised more than it did. It
now also drops every remaining `<` and `>`.

validateOrigin looked for markup in the sanitized string, after the tags
had been removed, so `<b>x</b>` passed as a valid origin. It now checks
the raw origin first.

pr-any.yml gets permissions: contents: read and lock.yml the
issues/pull-requests/discussions write it needs (CodeQL
actions/missing-workflow-permissions #3, #23).

* test(ui): typed cases and the matchers the test shim implements

The first version used it.each (untyped rows, @typescript-eslint/no-unsafe-argument) and expect().not.toMatch, which @pezkuwi/dev-test does not implement. Plain loops over string[] and a boolean assertion instead. With the repo runner: 5/5 pass on the fix, 3 of 5 fail on main's sanitizeOrigin.ts.

.github/workflows/lock.yml

@@ -5,6 +5,12 @@ on:
   schedule:
     - cron: '20 2/3 * * *'
 
+
+permissions:
+  issues: write
+  pull-requests: write
+  discussions: write
+
 jobs:
   lock:
     runs-on: ubuntu-latest

.github/workflows/pr-any.yml

@@ -1,6 +1,10 @@
 name: PR
 on: [pull_request]
 
+
+permissions:
+  contents: read
+
 jobs:
   pr:
     strategy:

packages/extension-ui/src/util/sanitizeOrigin.spec.ts

@@ -0,0 +1,45 @@
+// Copyright 2019-2026 @pezkuwi/extension-ui authors & contributors
+// SPDX-License-Identifier: Apache-2.0
+
+import type * as _ from '@pezkuwi/dev-test/globals.d.ts';
+
+import { sanitizeOrigin, validateOrigin } from './sanitizeOrigin.js';
+
+// The origin a site claims for itself is attacker-chosen text.
+const WITH_MARKUP: string[] = [
+  '<scr<script>ipt>alert(1)</script>',
+  '<<b>img src=x onerror=alert(1)>',
+  'a<b',
+  'x>y'
+];
+
+// validateOrigin used to look for markup in the sanitized string, after the
+// tags it was looking for had been removed: `<b>x</b>` passed as valid.
+const FLAGGED: string[] = ['<scr<script>ipt>evil', '<b>x</b>'];
+
+const ACCEPTED: string[] = ['https://app.pezkuwichain.io', 'http://localhost:3000', 'Pezkuwi Apps'];
+
+describe('sanitizeOrigin', () => {
+  it('keeps a normal origin as it is', () => {
+    expect(sanitizeOrigin('Pezkuwi Apps')).toEqual('Pezkuwi Apps');
+  });
+
+  // One pass of tag removal turns `<scr<script>ipt>` into `<script`.
+  it('leaves no angle bracket behind', () => {
+    for (const origin of WITH_MARKUP) {
+      expect(/[<>]/.test(sanitizeOrigin(origin))).toBe(false);
+    }
+  });
+
+  it('flags an origin that carried markup', () => {
+    for (const origin of FLAGGED) {
+      expect(validateOrigin(origin)).toBe(false);
+    }
+  });
+
+  it('accepts ordinary origins', () => {
+    for (const origin of ACCEPTED) {
+      expect(validateOrigin(origin)).toBe(true);
+    }
+  });
+});

packages/extension-ui/src/util/sanitizeOrigin.ts

@@ -22,6 +22,11 @@ const sanitizeHtml = (input: string, options?: {
       .replace(/<[^>]*>/g, ''); // Remove other HTML tags
   }
 
+  // Tag removal is a single pass, so nested input such as `<scr<script>ipt>`
+  // leaves a new tag behind. Drop every remaining angle bracket; an origin never
+  // contains one, and validateOrigin already treats them as suspicious.
+  sanitized = sanitized.replace(/[<>]/g, '');
+
   // Normalize whitespace
   sanitized = sanitized
     .trim()
@@ -44,6 +49,11 @@ export const sanitizeOrigin = (origin: string): string => {
 
 // Validate that the origin looks legitimate
 export const validateOrigin = (origin: string): boolean => {
+  // Markup is judged on what the site sent: the sanitized form no longer has any.
+  if (/[<>]/.test(origin)) {
+    return false;
+  }
+
   const sanitized = sanitizeOrigin(origin);
 
   // Special case: always allow localhost URLs
@@ -53,7 +63,6 @@ export const validateOrigin = (origin: string): boolean => {
 
   // Check for suspicious patterns
   const suspiciousPatterns = [
-    /[<>]/, // HTML characters
     /^\s*$/, // Empty or whitespace only
     /.{50,}/, // Extremely long names
     /^(?!https?:\/\/).*:\/\//, // Any protocol other than http/https