kube-mcp

The server directly constructs and executes `kubectl` commands using `child_process.exec` where many arguments (e.g., pod names, commands for `exec`, file paths for `apply` and `cp`, patch data for `patch`, labels/annotations) are derived directly from user input without apparent sanitization. This is a severe command injection vulnerability, allowing a malicious actor to potentially execute arbitrary code on the host system where the MCP server is running. For example, injecting shell metacharacters into the `command` argument of the `exec` tool, or the `file` argument of `apply`, could lead to Remote Code Execution. The `port-forward` tool also has the potential to expose internal services if misused. Running this server requires significant trust in the input it receives and robust sanitization of all user-controlled arguments is critical.