Kinds of causation:
Although causation by induction seems the least sound of the three, the other kinds are based on it. We are confident that balls behave in a certain way because we have seen them interact and have played with them. We also seem to have innate intuitions about causal interactions between solid objects. Even babies know that solid objects don't pass through one another.
But why is the ball solid? Well, the ball is only solid under a limited range of conditions. Too hot and the ball will melt or vaporize. Solidness is an emergent property of the interactions of a large number of atoms. In a certain temperature range the ball is solid because of the interactions between electrons in the atoms. On one view, an atom is mostly empty space. Electrons don't actually bounce off each other, the electromagnetic force between them is mediated by the exchange of virtual photons. But that wouldn't happen if electrons didn't have electric charge. Why do they? So far as we know, they just do. It's an induction. Physics has given us some insight into why objects can be solid or liquid, but we still don't have a complete casual mechanism free from inductive assumptions.
So in general, examining a causal mechanism at a higher level of detail can give us some insight into when the mechanism might break down, but it never eliminates all doubt. That is, we learn something about the nature of “all other things”.
Agent causation relies on the particularly unsound process of inferring intention. Usually we see intention as a subjective mental state, but we freely infer intention based on actions and assumed motivations, and we tend to see our inferred intentions as being more reliable than the reported subjective intentions of the other.
This is in fact reasonable on several levels, even leaving aside the social level explanation “he's lying”. We can act in an intentional way without forming a conscious intention (catching a ball unexpectedly thrown at us), and most animals probably act this way all the time. There is a good argument that our intention is actually generated unconsciously, based on weighing countless factors, most of which we never consciously consider. For this reason, we don't accurately know our own intention.
The intuitive “cause of existence” is like the ideal of a sufficient cause, but on close examination we see that this sort of cause isn't actually knowable at all. All we can do is change some things while trying to keep everything else equal. We can never specify what “everything else” is in full detail. This is as true in counterfactual interpretations as in experimental manipulation or statistical inference. In an experiment the control is supposed to capture “everything else”.
What is the “cause of existence” of a marble?
The idea of sufficient cause is driven by the social need to assign blame and award praise. Once you've found sufficient cause for an event, then you can identify all the culprits.
These social meanings of causation related to division of labor and moral responsibility are quite different from the pragmatic mechanisms of a manipulation theory of causation. We can reason in a precise way about necessary causes (preconditions) and can often convincingly demonstrate that a causal connection is real. This sort of understanding begins with being constantly on the watch for meaningful coincidences (correlations). A coincidence doesn't prove a causal connection, but making a change in the world (an experimental manipulation) can.
Because of the big payoff from being able to manipulate the world, we are highly attuned to noticing coincidences and are motivated to come up with and to act on causal stories. It is easy to fall for the narrative fallacy, so skepticism is always called for. Scientists are human, and their theories aren't exempt. Yet skepticism has no inherent limit; as the a ancient Greek skeptics understood, the critic never has to concede defeat, while a theory's supporter can be defeated.
There is a fundamental asymmetry in knowledge. Things can be proven false, but (outside of purely artificial worlds such as mathematics) nothing can ever be proven true, and indeed the impossibility of specifying “all other things” means that real world causal rules are prone to unpredictable breakdowns. Skepticism is infinite and rigorously founded; causal knowledge is fragile, scarce and valuable.