Scripting vs. RMM policies: where do you draw the line?

For routine workstation hygiene (disk cleanup, local admin pruning, agent health), what do you implement as portable scripts vs. native RMM policy objects? Post a small snippet or policy export and explain why it lives there.