Mediation without hypothesizing a direct effect

I know the current view (contrary to Baron & Kenny) is that you do not need to find a significant direct effect to test for mediation (e.g., Shrout & Bolger). However, can you still test for mediation if you do not hypothesize a direct effect?

For example, I'm testing a model in which X predicts Y and Y predicts Z (X --> Y --> Z), but I do NOT hypothesize that X will be related to Z. Can you still test the indirect effect from X to Z through Y and call it mediation? Or is it more of a cascading effects model?

Thank you.