The vulnerability CVE-2025-47913 exists in the golang.org/x/crypto/ssh/agent package used by Red Hat Trusted Artifact Signer (RHTAS). When the SSH agent client receives a generic SSH_AGENT_SUCCESS (0x06) message in response to requests expecting typed replies, the unmarshalling logic triggers a panic due to unhandled message types. This leads to process termination via panic("unreachable") or nil-pointer dereference. The flaw allows a malicious or misbehaving SSH agent or forwarded connection to cause a denial of service by crashing the client process. The issue is rooted in unsafe assumptions in the unmarshalling code, making it a design-level reliability problem rather than a simple error-handling bug. The vulnerability affects RHTAS versions 1.3.0 through 1.3.2 and is rated as important due to its impact on availability. Red Hat has not yet provided a patch or mitigation that meets their standards.

This vulnerability can be exploited by any malicious or misbehaving SSH agent to forcibly crash the client process that interacts with the agent. This results in a denial of service condition, which can disrupt critical automation, developer tooling, or CI/CD pipelines relying on SSH agent interactions. The attack requires no privileges or user interaction and can be triggered remotely if SSH agent forwarding or shared sockets are used. There is no impact on confidentiality or integrity, but availability is significantly affected.

Red Hat Product Security currently states that no mitigation is available that meets their standards for usability, deployment, applicability, or stability. Users should monitor Red Hat advisories for future updates or fixes. Until a patch is released, consider limiting exposure of SSH agent sockets to untrusted users or networks and avoid forwarding SSH agents in untrusted environments where possible.