IF instructions are required (and I find it isn't often if the overall course design is well thought through and intuitive), I opt for the time-delayed option on some of the earlier screens. I usually do this by having a relatively long time set on the base of the slide (e.g. 30 or 45 seconds) and having a trigger that shows an instruction layer at 'end of timeline'. This way, if the learner does what is required within the 30 or 45 seconds and moves off the screen they don't see the instruction layer. If they take too long on the screen (or perhaps take a wrong action), they do.
The best games take this kind of approach. Rarely are instructions explicit; players (learners, in our case) are given credit for being able to take appropriate action and get prompted with some kind of instruction only if and when it looks like it's necessary.