Each task comes from a different team with its own namespace, image and names. Nothing tells you which command to use: read the request, look at what is running, then act. The checker reads the cluster and the files you write, so any approach that produces the requested result passes. Watch the namespace in every task; most wrong answers here are right commands pointed at the wrong place.
A worked pattern on the first task's data.