Excalidraw CVE-2026-97224

Exploitation requires a victim to import an attacker-supplied Excalidraw file (or otherwise load scene data) whose customData.generationData.html field carries an HTML/script payload processed by the restore.ts import handler in Excalidraw <=0.18.1. The key limiting factor is required user interaction (UI:P): the attacker cannot trigger execution without the victim opening/importing the crafted file; there is no authentication requirement (PR:N). Impact is confined to client-side script execution in the Excalidraw page origin, and no server-side or availability impact is indicated.

This is a low-severity, high-caveat issue. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:P) describes network-reachable, unauthenticated exploitation but requires user interaction (UI:P) - specifically a victim importing an attacker-crafted Excalidraw file - and the only impact is VI:L (limited integrity), with VC:N and VA:N, yielding VulDB's 2.1 score. My own read is slightly higher: DOM-based XSS executing in the origin can read/alter page content, so a C:L/I:L profile (CVSS 3.1 ~5.4-6.1 with S:C) is defensible, but this remains a low-to-medium real-world priority, not a critical one. SSVC data is not provided. Exploitation status: publicly available exploit code exists (VulDB states 'The exploit is now public'), but this is NOT active exploitation - there is no CISA KEV listing and no EPSS score in the input, so treat POC availability and in-the-wild exploitation as distinct signals. The vendor did not respond to disclosure, so no official patch is confirmed. Net: a genuine but low-impact client-side XSS gated on user-initiated file import; prioritize patching where Excalidraw processes untrusted imported files, but it is not an emergency.

An attacker crafts a malicious .excalidraw file with a poisoned customData.generationData.html value and shares it (via a link, chat, or shared-workspace) with a victim. When the victim imports/opens the file in an affected Excalidraw instance (<=0.18.1), the restore.ts import handler renders the attacker HTML, executing script in the origin. A public exploit is available, lowering the skill barrier, though success depends on the victim performing the import action (UI:P).