The exciting part of this discovery is that any sequence can be used to produce the crRNA in the lab (it doesn’t have to be viral DNA), and that this system can be made to work in any cell type (not just bacteria). The only limitation is that the crRNA needs to be relatively short (27 to 47 nucleotides/letters long) and specific to the target DNA sequence (the sequence you want to cut). Thus scientists can target and cut any gene they want.

Source: www.nbri-nju.com
Specificity is accomplished by ensuring that the selected sequence is relatively long and non-repetitive. Think of sequence specificity like this:
If you have a genome composed of a random sequence of A-s, T-s, G-s and C-s then the chances of finding two regions that are complementary to ATGGCT is greater than ATGGCTGGCTAGTCG. Thus the longer sequence, the less likely it is that it will be found in two different places in the genome and therefore your chosen sequence is less likely to bind to a gene you do not want to modify.