OpenClaw versions before 2026.3.7 include a flaw in the fetchWithSsrFGuard function that fails to filter custom authorization headers when following redirects across another origin. As a result, a redirect that points to a different domain can cause the original request’s headers such as X-Api-Key or Private-Token to be forwarded to the attacker’s server. This exposure leaks credentials that are otherwise intended only for the target destination. The weakness maps to memory data protection failures (CWE‑522) and would allow an attacker to read sensitive authentication tokens.

Affected software is the OpenClaw web framework, all releases earlier than 2026.3.7, running on a Node.js environment. The vulnerability is tied to the OpenClaw package as identified by the CNA. Users deploying these versions should check their installed package version; any earlier release is vulnerable.

The flaw carries a high severity CVSS score of 8.8 but has a very low probability of exploitation with an EPSS below 1 %. It is not listed in the CISA KEV catalog. Based on the description, it is inferred that an attacker can trigger a cross‑origin redirect by manipulating input to fetchWithSsrFGuard. Although the conditions for exploitation are not trivial, the potential impact of credential theft makes the risk significant. Until a patch is applied the vulnerability remains a credible threat.