← AI agent vulnerabilities

CVE-2026-47250

mcp-server-kubernetes

6.1
Severity
Medium
Published
June 11, 2026
Last modified
June 17, 2026

mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.7.0, the kubectl_generic tool in mcp-server-kubernetes passes user-supplied flags directly to kubectl without any allowlist, enabling a privilege escalation attack within Kubernetes environments. An attacker who already has limited cluster or codebase access, for example, a developer with pod-deployment permissions but not cluster-admin credentials, can plant a single structured JSON line in an application's log output. When an operator with a privileged kubeconfig uses the MCP server to read those logs and their AI agent follows the injected instruction, kubectl_generic is called with --server=https://attacker.example.com and --insecure-skip-tls-verify=true. kubectl sends all API requests, including the Authorization: Bearer <token> header from the operator's kubeconfig to the attacker's endpoint. The captured token can then be replayed directly against the real Kubernetes API server, granting the attacker the full RBAC permissions of the operator's service account. This issue has been patched in version 3.7.0.

CVSS vector CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:H/I:N/A:N

Weakness

References

CVE-2026-47250 on NVD

Listed in

MediumMCP servers

More mcp-server-kubernetes CVEs

Jun 11
CVE-2026-465198.8

Mcp-server-kubernetes exposes three environment variables (ALLOW_ONLY_READONLY_TOOLS, ALLOW_ONLY_NON_DESTRUCTIVE_TOOLS, ALLOWED_TOOLS) documented as access controls for restricting which Kubernetes…

Apr 15
CVE-2026-398848.3

Versions 3.4.0 and prior contain an argument injection vulnerability in the port_forward tool in src/tools/port_forward.ts, where a kubectl command is constructed via string concatenation with…