Can I create a simple commitment scheme using a secure hash function?
If so, is concatenation with a random secret enough to preserve hiding? (i.e. $C = H( random\_string || message)$)
Yes. If you publish such a commitment. And you model the hash as a random function it willl not only be preimage resistant but there will be many possible pairs of random string and message which will match the commitment. If the random string is as big as the hash output most possible message values can produce the commitment for some random string. So even an attacker with infinite compute power will not be able to consistently discover the message, while an attacker with bounded computing power won't be able to learn anything about the message.
When the commitment is revealed, we know the attacker didn't cheat because collision resistance means the committer (With bounded conputing resources) won't be able to produce a commitment which matches two distinct known messages.