Thanks! In general you shouldn't need to do this, even if the job was a mistake. Kubernetes cleans up the job automatically a week after it terminates -- whether it completed or failed -- and there's nothing wrong with leaving it until then.

We use mwscript-cleanup as a systemd timer to additionally tidy up Helm secrets (associated with the implementation detail that each job is a new Helm release) but that's all hidden plumbing; I don't think it's what you care about.

If it's important to delete the job ahead of schedule, the best way is probably just a kubectl delete command:

$ kube_env mw-script-deploy eqiad  # Carefully use the -deploy user to get delete privileges
$ kubectl get job -l username=lucaswerkmeister-wmde  # Look up the job name, if you don't have it (mwscript-k8s prints it now)
$ kubectl delete job mw-script.codfw.r72z2aop

That automatically deletes the associated pod as well, terminating it if it's still running. It doesn't do anything with the Helm secret, but mwscript-cleanup still will when it runs later.