Critical XSS Flaw Hits OpenClaw Dashboard

A critical stored cross-site scripting (XSS) vulnerability, tracked as CVE-2026-66421, has been identified in the OpenClaw Dashboard. This flaw allows unauthenticated remote attackers to execute arbitrary JavaScript within an administrator's browser session by injecting malicious HTML into agent transcript messages.

What's at Risk

The vulnerability exists within the sessions API of the OpenClaw Dashboard. Because the application fails to properly sanitize input before rendering it via innerHTML on the default landing page, attackers can inject payloads such as img tags with onerror attributes. This puts organizations using the OpenClaw Dashboard at risk of full session token theft and unauthorized execution of administrative commands, including the modification of sensitive agent instruction files.

Systems that are exposed to the internet are at the highest risk, as the attack requires no prior authentication to the dashboard. Any environment where administrative sessions are active is potentially vulnerable to exploitation if an attacker can submit a crafted message through the affected API.

How the Flaw Works

Stored XSS is a type of injection attack where a malicious script is permanently stored on a target server, such as in a database or message log. When a user—in this case, an administrator—views the stored data, the browser interprets the malicious input as legitimate code rather than plain text. This allows the script to run within the context of the user's active session.

When this occurs, the script gains the same permissions as the logged-in user. Attackers typically use this access to hijack session cookies, perform actions on behalf of the user, or exfiltrate sensitive data displayed on the page. Because the browser trusts the origin of the script, security controls like the Same-Origin Policy are bypassed, making this a highly effective method for gaining unauthorized access to administrative interfaces.

How to Protect Your Systems

• Review and audit all recent agent transcript logs for suspicious HTML or script tags.
• Restrict access to the OpenClaw Dashboard sessions API to trusted network segments or VPN-only access.
• Implement or update Content Security Policy (CSP) headers to prevent the execution of inline scripts and unauthorized external resources.
• Monitor administrative accounts for unusual activity or unauthorized configuration changes to agent instruction files.
• Apply vendor-provided patches as soon as they are available to remediate the unsanitized rendering of input.

With a CVSS score of 9.3, this critical vulnerability presents a significant risk to the integrity and confidentiality of the OpenClaw Dashboard. Given that the flaw allows for unauthorized administrative actions, administrators should prioritize securing their environments immediately. Promptly addressing such flaws is the most effective way to prevent session hijacking and the subsequent compromise of sensitive system configurations.