Marko is a declarative, HTML-based language for building web apps. Prior to marko version 5.38.36 and prior to @marko/runtime-tags 6.0.164, when dynamic text is interpolated into a <script> or <style> tag the Marko runtime failed to prevent tag breakout when the closing tag used non-lowercase casing. An attacker able to place input inside a <script> or <style> block could break out of the tag with </SCRIPT>, </Style>, etc. and inject arbitrary HTML/JavaScript, resulting in cross-site scripting. This issue has been patched in marko version 5.38.36 and @marko/runtime-tags 6.0.164.
Marko is a declarative, HTML-based language for building web apps. Prior to marko version 5.38.36 and prior to @marko/runtime-tags 6.0.164, when dynamic text is interpolated into a <script> or <style> tag the Marko runtime failed to prevent tag breakout when the closing tag used non-lowercase casing. An attacker able to place input inside a <script> or <style> block could break out of the tag with </SCRIPT>, </Style>, etc. and inject arbitrary HTML/JavaScript, resulting in cross-site scripting. This issue has been patched in marko version 5.38.36 and @marko/runtime-tags 6.0.164.
Marko is a declarative, HTML-based language for building web apps. Prior to marko version 5.38.36 and prior to @marko/runtime-tags 6.0.164, when dynamic text is interpolated into a <script> or <style> tag the Marko runtime failed to prevent tag breakout when the closing tag used non-lowercase casing. An attacker able to place input inside a <script> or <style> block could break out of the tag with </SCRIPT>, </Style>, etc. and inject arbitrary HTML/JavaScript, resulting in cross-site scripting. This issue has been patched in marko version 5.38.36 and @marko/runtime-tags 6.0.164.
### Summary
When dynamic text is interpolated into a `<script>` or `<style>` tag the Marko runtime failed to prevent tag breakout when the closing tag used non-lowercase casing.
An attacker able to place input inside a `<script>` or `<style>` block could break out of the tag with `</SCRIPT>`, `</Style>`, etc. and inject arbitrary HTML/JavaScript, resulting in cross-site scripting.
### Details
The affected helpers used case-sensitive regular expressions to detect attempts at closing the surrounding tag:
```js
// packages/runtime-tags/src/html/content.ts
const unsafeScriptReg = /<\/script/g;
const unsafeStyleReg = /<\/style/g;
// packages/runtime-class/src/runtime/html/helpers/escape-script-placeholder.js
const unsafeCharsReg = /<\/script/g;
// packages/runtime-class/src/runtime/html/helpers/escape-style-placeholder.js
const unsafeCharsReg = /<\/style/g;
```
HTML tag names are case-insensitive in the browser parser, so inputs such as `</SCRIPT>`, `</Script>`, or `</sTyLe>` were not matched by these regexes and passed through the helpers unchanged. A browser rendering the output treats the mixed-case end tag as a valid closing tag, terminating the script or style context, and then parses anything that follows as HTML.
The Marko compiler routes interpolated values inside `<script>` and `<style>` tags through these helpers automatically (see `native-tag.ts:1080-1085`), so application code following the framework's conventions had no way to detect or compensate for the gap.
### PoC
```marko
$ const userCode = "</SCRIPT><script>alert(1)//";
<script>
const data = ${JSON.stringify(userCode)};
</script>
```
Would yield the following:
```html
<script>const data = "</SCRIPT><script>alert(1)//";</script>
```
Which is then parsed in any WHATWG-compliant browser as:
```html
<script>const data = "</script>
<script>alert(1)//";</script>
```
### Impact
Cross-site scripting. Any Marko template that explicitly interpolates untrusted data inside a `<script>` or `<style>` block is affected.
Stored XSS is trivial if the value originates from any persisted user input (username, profile bio, comment body, etc.) that is later embedded in a script tag during rendering. Exploitation yields arbitrary JavaScript execution in the victim's browser, enabling session token theft, account takeover, and arbitrary actions as the victim.
Since the internal `_escape_script` and `_escape_style` helpers are the framework's designated defense against script/style tag breakout, applications following standard Marko patterns had no obvious reason to add a second layer of sanitization.
This does not affect scripts or hydration state serialized by Marko itself — only templates that explicitly interpolate untrusted values inside a <script> or <style> tag.
### Patch
Commit `19d4b37d0` — `fix: html script, style, and comment escaping`.
```diff
- const unsafeScriptReg = /<\/script/g;
+ const unsafeScriptReg = /<\/script/gi;
- const unsafeStyleReg = /<\/style/g;
+ const unsafeStyleReg = /<\/style/gi;
```
The same commit also introduced an `_escape_comment` helper and corresponding `escape-comment-placeholder.js`, hardening HTML comment escaping as a related preventative fix. Test fixtures were added under `escape-script-case`, `escape-style-case`, and `escape-comment`.
### Workarounds
Upgrade to the patched release. As a short-term mitigation on affected versions, pre-sanitize any untrusted data before it reaches a template position rendered inside a `<script>` or `<style>` tag — e.g. normalize `</script`, `</style`, and their mixed-case variants before interpolation, or avoid direct interpolation of untrusted values inside these tags entirely.
EUVD source date text has no stated timezone and is retained without inventing a UTC instant. Related bulletin membership is not an alias, common cause or applicability claim. A missing vector with a zero score is treated as unrated. Complete vendor/product and underlying source evidence remains in the original JSON.
Original publisher date fields
datePublished: May 8, 2026, 3:22:50 PM
dateUpdated: May 8, 2026, 9:26:58 PM
No normalized affected-package data is projected for this source revision. The original publisher record may contain additional product and version evidence; this does not establish that a package is unaffected.
NVD analysis and CPE criteria are source context, not proof that a package or your environment is affected. Complete affected data, logical configurations and SSVC are retained in the original JSON.
The original JSON retains the logical configuration operators, exclusions and version bounds.
No normalized affected-package data is projected for this source revision. The original publisher record may contain additional product and version evidence; this does not establish that a package is unaffected.