You know how to apply the rule of addition because it is an algorithm spelled out beforehand. This is not the kind of rule Wittgenstein had in mind.
So Kripke's example is a bit silly. Kripke's claim is that "That there is no fact about your past usage of the addition function that determines you have the right answer." But the rule is stated. You do have evidence that you are meant to use the ordinary rules of addition in this case. You have been told they are correct by authorities in your culture. You know when you are following the rules because you have been inaugurated into the rules with extensive operant conditioning.
A better example is how children acquire grammar. No one sits them down and states the rules, because the rules would have to be stated in some language, and the child is just beginning to acquire their first language.
Wittgenstein used this acquisition of first language as a general model for all learned meaning. He had in mind that the rules of a culture, a science, or a philosophy are really the rules of a 'language game', which are not expressed beforehand by anyone, and are never clearly expressible by anyone, because they are subject to continual change. He imagines that people decide what rules they like and converge together on the right answers, praising the people who guess the 'right' answer first or more often. Then the actual statements of cultural, scientific or philosophical thoughts are mere approximation to this more genuine set of rules no one can formulate completely.
The paradox is that there is no way to discern a rule by simple observation when you cannot state them clearly. In that case, you could never really learn language without Chomsky's 'built-in grammar instinct' that already knows the limits on what is possible valid grammar because grammar itself is evolved to have boundaries.
Kripke says Wittgenstein is wrong (and Chomsky closer to correct), because there is no real way to learn or agree on a broad rule with a finite number of examples, unless you state it. Wittgenstein imagines, for instance, that grammar should be learned by examples alone, and Kripke would protest that no matter how long you spoke a language, unless you actually stated a rule of grammar, you would never be truly sure you are right.
But people have languages with no recorded grammar, and their children do acquire them. At the same time, folks doubt Chomsky, now that we know dogs learn complex language rules like the process of elimination, and apes can learn sign-language. Kripke's argument that it is not possible is logical, but does not seem to matter. There has to be a way learning rules works, that is not just about passively collecting examples.
The way out of this, to my mind is Popper. You can generally be really sure when you are wrong, and you can learn enough about the world from how seriously you are corrected that you can come to follow the rule almost exactly after relatively few tries. This is verified by interactive simulations of evolution that led us to discover genetic algorithms. A good learning algorithm can converge on something very close to a goal after relatively few clear corrections or competitive pairings.
So I would consider Wittgenstein vindicated on this one. It does not matter than induction or abduction does not really work, competition still allows for ways to negotiate rules over time with arbitrary accuracy.