To follow a rule (e.g. ‘under circumstances \(C\), do \(A\)’, or ‘Whenever you see a cat, raise your hand’), we have to understand the concepts that are involved within (i.e. \(C\) or cat, or raising one’s hand).
However, to grasp a concept (such that the original rule can be followed) requires a further rule.
Put another way, because rules have many interpretations, for any rule \(R\) we need another rule which tells us whether or not we correctly applied \(R\).
If a law says ‘Every man must serve in the army’, then it will naturally require a law for determining who qualifies as ‘man’. That law (say, ‘A man is whatever the scientific experts label with the word man’) will naturally require a law for determining who qualifies as ‘scientific expert’… the regress will continue if we try to adjudicate ‘scientific expert’ with another rule.
This skepticism about rules is really a skepticism about a certain theory of meaning (i.e. semantic skepticism).