Shellbox, as deployed in production, effectively requires k8s to manage the multiple containers it needs (php-fpm, httpd, prometheus exporters for both).
We can currently catch regressions in Shellbox with the k8s staging cluster in prod, but we have no active integration testing of the MediaWiki Shellbox client library against a running Shellbox until it hits prod.
Short of having a k8s cluster in beta, it theoretically could be possible to use minikube to set up a running Shellbox, but likely all of that would be unpuppetized.