Executive brief
OpenClaw Dashboard, a monitoring tool for AI agents, contains a security flaw that allows unauthorized attackers to take control of the system. By submitting a specially crafted username during a failed login attempt, an attacker can plant malicious code that executes when an administrator views the system's audit logs. This could lead to full account takeover, allowing the attacker to modify agent configurations and sensitive system settings.
Technical details
A stored cross-site scripting (XSS) vulnerability exists in OpenClaw Dashboard v3.0.0 due to improper neutralization of input in the audit logging mechanism. An unauthenticated attacker can send a crafted username containing malicious script via a failed login POST request. The application records this input verbatim in the audit log. When an administrator views the notification panel, the application renders the log entry using innerHTML. Because the Content-Security-Policy (CSP) is overly permissive and allows inline event handlers, the injected script executes in the context of the administrator's session. This allows the attacker to perform actions on behalf of the administrator, such as editing agent instruction files or changing system configurations. As of the advisory date, the vulnerability affects version 3.0.0.
Affected products
- tugcantopaloglu openclaw-dashboard 3.0.0
Timeline
- 2026-07-30: advisory: Initial disclosure by VulnCheck and NVD