GHSA-MXM6-V9R6-R94C

Vulnerability from github – Published: 2026-09-16 22:14 – Updated: 2026-09-16 22:14
VLAI
Summary
@nuxtjs/mdc's URL sanitizer misses SVG xlink:href and data:text/html, allowing XSS from untrusted markdown at the default configuration
Details

Summary

@nuxtjs/mdc renders untrusted markdown (including raw HTML) to a Vue component tree. Across two prior advisories it added a URL/attribute sanitizer to block dangerous links in that HTML: validateProps / validateProp and an unsafeLinkPrefix deny-list (dist/runtime/parser/utils/props.js). The sanitizer runs at parse time (dist/runtime/parser/compiler.js) and parseMarkdown enables raw HTML by default (allowDangerousHtml: true, dist/runtime/parser/options.js), so the sanitizer is the only barrier and it applies with no configuration required.

Two sibling vectors bypass that sanitizer at the default configuration:

  1. SVG anchor xlink:href. validateProp only scheme-checks attributes named exactly href or src:

if (attribute === "href" || attribute === "src") return isAnchorLinkAllowed(value); return true;

An xlink:href (parsed to the hast property xLinkHref) is neither, so a javascript: URL on an SVG <a> is passed through. The renderer maps the property back to the real attribute (MDCRenderer.vue: find(html, "xLinkHref").attribute is xlink:href), so the output element is <a xlink:href="javascript:...">. Clicking it runs the script in the page origin. Plain <a href="javascript:..."> is correctly stripped, which is what makes this the un-patched sibling.

  1. <iframe src="data:text/html,...">. data:text/html is present in unsafeLinkPrefix, but the check compares it against url.protocol:

if (unsafeLinkPrefix.some((prefix) => url.protocol.toLowerCase().startsWith(prefix))) return false;

For any data URI url.protocol is just "data:", so "data:".startsWith("data:text/html") is always false. Every data:text/* entry in the deny-list is therefore dead code, and <iframe src="data:text/html,<script>...</script>"> is allowed (iframe is not in the render-time dangerousTags, which is only ["script","base"]). The framed document executes script in an opaque origin. For contrast, srcdoc and object are blocked, so this is a precise gap rather than a general absence of filtering.

Reproduction

I will attach the zip file for POC, you can simply extract and run ./poc.sh to install mdc and show the poc in the html file. nuxtjs-mdc-xss_poc.zip

Two zero-argument checks:

  1. sh poc/poc.sh installs @nuxtjs/mdc and runs parseMarkdown (the documented API) at default. It shows the parsed tree retains a { xLinkHref: "javascript:..." } and iframe { src: "data:text/html,..." }, while the control payloads href="javascript:..." and srcdoc=... are removed by the sanitizer. This isolates the sanitizer bypass deterministically.
  2. poc/poc.sh also serves poc/poc.html over http (data: iframes and javascript: links are restricted under the file:// origin, so http is used). Open the printed URL and click the blue SVG link. The page contains the exact DOM the renderer produces for those parsed nodes; clicking the SVG link executes script in the page origin (same-origin), and the data:text/html iframe executes on load. The page prints VULNERABLE for each that fires.

Both vectors were confirmed executing in a current Chromium build: the SVG xlink:href link runs script in the document origin on click, and the data:text/html iframe runs script on load.

Suggested fix

In validateProp, scheme-check xlink:href (and the hast xLinkHref) the same way as href/src. In isAnchorLinkAllowed, compare the dangerous MIME-typed entries against the full URL (or href), not against url.protocol, so data:text/html is actually matched; or add iframe to the render-time dangerous-tag set / restrict iframe src schemes.

Severity
8.1 (High)
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "npm",
        "name": "@nuxtjs/mdc"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.22.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-63671"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-184",
      "CWE-79"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-09-16T22:14:01Z",
    "nvd_published_at": "2026-09-16T15:17:40Z",
    "severity": "HIGH"
  },
  "details": "## Summary\n\n`@nuxtjs/mdc` renders untrusted markdown (including raw HTML) to a Vue component tree. Across two prior advisories it added a URL/attribute sanitizer to block dangerous links in that HTML: `validateProps` / `validateProp` and an `unsafeLinkPrefix` deny-list (`dist/runtime/parser/utils/props.js`). The sanitizer runs at parse time (`dist/runtime/parser/compiler.js`) and `parseMarkdown` enables raw HTML by default (`allowDangerousHtml: true`, `dist/runtime/parser/options.js`), so the sanitizer is the only barrier and it applies with no configuration required.\n\nTwo sibling vectors bypass that sanitizer at the default configuration:\n\n1. SVG anchor `xlink:href`. `validateProp` only scheme-checks attributes named exactly `href` or `src`:\n\n   ```\n   if (attribute === \"href\" || attribute === \"src\") return isAnchorLinkAllowed(value);\n   return true;\n   ```\n\n   An `xlink:href` (parsed to the hast property `xLinkHref`) is neither, so a `javascript:` URL on an SVG `\u003ca\u003e` is passed through. The renderer maps the property back to the real attribute (`MDCRenderer.vue`: `find(html, \"xLinkHref\").attribute` is `xlink:href`), so the output element is `\u003ca xlink:href=\"javascript:...\"\u003e`. Clicking it runs the script in the page origin. Plain `\u003ca href=\"javascript:...\"\u003e` is correctly stripped, which is what makes this the un-patched sibling.\n\n2. `\u003ciframe src=\"data:text/html,...\"\u003e`. `data:text/html` is present in `unsafeLinkPrefix`, but the check compares it against `url.protocol`:\n\n   ```\n   if (unsafeLinkPrefix.some((prefix) =\u003e url.protocol.toLowerCase().startsWith(prefix))) return false;\n   ```\n\n   For any data URI `url.protocol` is just `\"data:\"`, so `\"data:\".startsWith(\"data:text/html\")` is always false. Every `data:text/*` entry in the deny-list is therefore dead code, and `\u003ciframe src=\"data:text/html,\u003cscript\u003e...\u003c/script\u003e\"\u003e` is allowed (iframe is not in the render-time `dangerousTags`, which is only `[\"script\",\"base\"]`). The framed document executes script in an opaque origin. For contrast, `srcdoc` and `object` are blocked, so this is a precise gap rather than a general absence of filtering.\n\n## Reproduction\n\nI will attach the zip file for POC, you can simply extract and run `./poc.sh` to install mdc and show the poc in the html file.\n[nuxtjs-mdc-xss_poc.zip](https://github.com/user-attachments/files/29175603/nuxtjs-mdc-xss_poc.zip)\n\nTwo zero-argument checks:\n\n1. `sh poc/poc.sh` installs `@nuxtjs/mdc` and runs `parseMarkdown` (the documented API) at default. It shows the parsed tree retains `a { xLinkHref: \"javascript:...\" }` and `iframe { src: \"data:text/html,...\" }`, while the control payloads `href=\"javascript:...\"` and `srcdoc=...` are removed by the sanitizer. This isolates the sanitizer bypass deterministically.\n2. `poc/poc.sh` also serves `poc/poc.html` over http (data: iframes and javascript: links are restricted under the file:// origin, so http is used). Open the printed URL and click the blue SVG link. The page contains the exact DOM the renderer produces for those parsed nodes; clicking the SVG link executes script in the page origin (same-origin), and the data:text/html iframe executes on load. The page prints VULNERABLE for each that fires.\n\nBoth vectors were confirmed executing in a current Chromium build: the SVG `xlink:href` link runs script in the document origin on click, and the data:text/html iframe runs script on load.\n\n## Suggested fix\n\nIn `validateProp`, scheme-check `xlink:href` (and the hast `xLinkHref`) the same way as `href`/`src`. In `isAnchorLinkAllowed`, compare the dangerous MIME-typed entries against the full URL (or `href`), not against `url.protocol`, so `data:text/html` is actually matched; or add `iframe` to the render-time dangerous-tag set / restrict iframe `src` schemes.",
  "id": "GHSA-mxm6-v9r6-r94c",
  "modified": "2026-09-16T22:14:01Z",
  "published": "2026-09-16T22:14:01Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/nuxt-content/mdc/security/advisories/GHSA-mxm6-v9r6-r94c"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-63671"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nuxt-content/mdc/pull/491"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nuxt-content/mdc/commit/61d636c2983f021288e4fc5c4006733b38cf0d53"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/nuxt-content/mdc"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nuxt-content/mdc/releases/tag/v0.22.1"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "@nuxtjs/mdc\u0027s URL sanitizer misses SVG xlink:href and data:text/html, allowing XSS from untrusted markdown at the default configuration"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.



Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.