In introductory discussions of Kant, and sometimes beyond, it is common to raise the problem of the level of generality for maxims. Kant's categorical imperative tells us,
Act only according to that maxim that you can at the same time will to be a universal law,
where a maxim is the rule you are making for yourself in making a decision. So, for instance, if I am considering whether I should make a promise that I know would have good consequences but that I also know I can never fulfill, my maxim is "Make a promise you know you will break if it gets good consequences", and Kant holds that I should consider whether this is something that can be consistently willed by a consistently rational being under all circumstances. The answer, of course, is that in making a promise you are relying on the norm of promises being something that are kept and yet you are also making an arbitrary exception for yourself. On Kant's view all immorality is a failure to treat moral principle as moral principle by making special exceptions for yourself or other people that do not derive simply and purely from what it is to be a rational being.
But, the line of thought also goes, what if we fiddle with our maxims? So one version is, suppose a person, call her Alice, is considering whether to repay a debt to another, call her Sara, because she doesn't want to do so. 'Don't pay your debts when you feel like not doing so' is not a universalizable maxim. But suppose Alice tinkers with her maxim, and says, "But my maxim is actually, 'don't pay your debts when you are Alice and you would have to be paying Sara'. This seems universalizable!"
Alice, however, is certainly wrong here, because she is lying about her maxim; the maxim being the actual subjective rule from which her action proceeds, her maxim in fact includes trying to pretend that her maxim is merely what she says it is in order to get the result she wants, and this is not universalizable. Deliberate maxim-tampering is not a problem with Kant, because it's the actual rule you are using to guide your action that needs to be considered, not the maxim that you arbitrarily and dishonestly claim that you are following. (We run into a similar bit of sophistry in discussions of double effect; people pretend that you can be intending something just by saying 'I am intending this, not that', but in fact your intentions are not declarations but your actual disposition and aim in acting. If you deliberately murder your enemy while saying, "I am only acting in self-defense," that does not mean you were acting in self-defense but that you are a liar as well as a murderer.) Maxims are not things put out by a maxim factory according to custom specifications. They are the rules actually in play.
But there are other attempts to raise the generality problem for maxims that don't involve any attempt at maxim-tampering. To use an example from Bergeron and Tramel's "Rightness as Fairness: Kant's Categorical Imperative", 'In order to make money, rob a bank', is not universalizable. But what if, from the beginning (so it is not maxim-tampering), the maxim is actually, 'In order to provide money for such-and-such person with such-and-such characteristics unique for me, rob such-and-such bank'. That is, suppose that this is literally my first and last thought, that I in particular should rob this bank in particular. If we blow this up on a universal scale, we get -- the one unique action.
In fact this is not true; as neither you nor the bank are a universal feature of rational beings, nor of the life of rational beings, and the maxim cannot be willed as something to which all rational beings in such a situation would be subject as rational beings (which is what Kant means by 'universal'), it is not universalizable at all. Merely saying you've universalized it is not true. Nor is it willable as universal law; there is nothing in it that could be willed as a law for every rational being at all. It is something you are explicitly willing for yourself as a unique individual, and it is contradictory to do that and also will it as something that should be applicable as a rule for everyone.
The problem with this kind of counterexample is that it is based on the false assumption that all that is required for a maxim to pass is to be implementable if made general. But Kant is very clear that the point of 'universality' is specifically that it must be the sort of thing that could apply to anyone.
In general, overspecification of maxim will get the same result, and thus will not pose a problem for the categorical imperative.
But could we go the other direction? To borrow another example from Bergeron and Tramel, suppose I am deciding whether to sit in my study, and the maxim that I adopt is, "In order to work comfortably, I should sit in the study." This doesn't seem universalizable. So surely something has gone wrong.
However, this is also not a problem for the categorical imperative. If this is literally my maxim, if we are not being loose in stating it, then it is an immoral maxim, because universalized it treats working comfortably as a decisive and definitive end for every rational being, which is not consistent with the concept of a being capable of acting according to universal law. In reality, we would generally assume that there are implicit conditions that are not explicitly stated here; as it stands it is a technical maxim, how to do something I might want to do (work comfortably), and does not actually specify the point of doing so. If I intend working comfortably as a universal moral end all rational beings must pursue, I am certainly a moral loon, and not rational at all. But if I intend working comfortably as the sort of thing I could do in these circumstances in order to be better prepared for doing my duty, this is an end that could be willed as an end of significance for all rational beings, and if that end universalizes, my real maxim is universalizable.
The problem with all of these is perhaps that people tend to assume that all maxims can be perfectly written down in simple sentences, but one of Kant's key points is that rules can often have conditions that are not explicitly stated. This is why there is only one categorical imperative. I can write down an imperative in a form that indicates no conditions at all -- "Do your homework" -- but any rational person would know that it in fact has conditions -- it would be a sign of irrationality to try to obey "Do your homework" even to the point of death, for instance, and it is not even the right kind of imperative to trump every other possible imperative that might come along. Likewise, your maxim may be stated as 'Do X for purpose Y', but Y itself might be only a thing being done for some other reason, and 'Do X' might be conditional on more general imperatives being met, and X's relation to Y might itself depend on some rational or irrational principle or other. All of these are relevant to whether the maxim can in fact be willed as universal law, that is, a rule of action a rational being could will for all rational beings.
There are problems that could be raised about Kant's whole apparatus of law and maxims, mostly to do with the assumptions he makes about how practical reason works, and how it relates to the will (Kant in fact rejects there being any significant difference between practical reason and will); but there is no problem with the levels of generality in maxims. What is relevant is your actual maxim, not some dummy maxim you state as a proxy, and it doesn't matter whether it is very specific or very general as long as it can be willed as universal law. Universalizability is not about big scale but about being the kind of thing appropriate to a rational being as such. Kant has all the resources he needs to handle the major classes of supposed counterexamples that constitute the generality problem for maxims.