A ** is just a pointer to a pointer. So where an instruction* contains the address of an instruction struct, an instruction** contains the address of an instruction* that contains the address of an instruction object.
To access the instruction pointed to by the pointer pointed to by an instruction**, you just use two asterisks instead of one, like (**p).repetitions or something similar.
You can visualize it like this:
instruction* ----> instruction
instruction** ----> instruction* ----> instruction
Remember, however, that simply declaring struct instruction** instructions; doesn't actually create an instruction struct. It just creates a pointer that holds a garbage value. You'll have to initialize it:
struct instruction inst;
// set members of inst...
*instructions = &inst;
...
(*instructions)->repetitions++; // or whatever
However, it looks like you're using an instruction** to point to an array of instruction*s. To initialize the array, you need a for loop:
instructions = malloc(sizeof(struct instruction*) * num_of_arrays);
for (i = 0; i < num_of_arrays; ++i)
instructions[i] = malloc(sizeof(struct instruction) * size_of_each_subarray);
And then you can access an element like instructions[i]->datamember.