CVE-2026-129236

Kubeshark MCP deployment arguments could select untrusted privileged images

kubeshark · Kubeshark MCP

First documented disclosure · Review checkpoint

Summary

Agent-controlled deployment input could change which images a privileged Kubernetes workload runs.

Severity and exploitation evidence

High · 7.2/10 · CVSS 3.1. Severity source.

Exploitation not established: Malicious exploitation status remains unverified in the reviewed sources.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

A positional argument was interpreted as a deployment option, crossing the tool’s intended argument boundary.

Conditions required

Destructive MCP tools are enabled and the host has permission to deploy Kubeshark workloads.

Potential impact

A selected privileged image can gain node access and cluster-administrator authority.

Affected versions and configurations

53.1.0 through versions before 53.4.0.

Fix and mitigation status

53.4.0.

Practical next steps

Upgrade to 53.4.0 or later and restrict destructive-tool deployment privileges.

Operating lesson

Treat deployment options as privileged inputs even inside a narrower tool.

Evidence and limits

Original disclosure 2026-08-13. The CVE Program published this record 2026-11-16.

Sources