Executive brief
LobeChat is an AI agent management and scheduling platform. A flaw in avatar upload handling allows authenticated users to inject malicious JavaScript code via SVG files, which executes when other users view the attacker's profile. This could lead to session hijacking, credential theft, or unauthorized actions taken on behalf of victims.
Technical details
A stored cross-site scripting (XSS) vulnerability exists in LobeChat through v2.2.13 in the user avatar upload functionality. An authenticated attacker can craft a malicious SVG file containing embedded JavaScript and upload it as their profile avatar. The vulnerability stems from insufficient input validation and output encoding of user-supplied SVG content. When other users view the attacker's profile or avatar, the stored malicious script executes in their browser within the application context, potentially allowing session hijacking, credential theft, or unauthorized actions. The vulnerability requires an authenticated user and user interaction (viewing the profile/avatar), but does not require elevated privileges.
Affected products
- LobeHub LobeChat through v2.2.13
Timeline
- 2026-08-10: disclosed