If in the set of natural numbers, all prime numbers $p$ have only two divisors, $1$ and $p$, and all composite numbers have at least three divisors, then can we also use these definitions for the set of integers to say that all negative numbers are composite numbers except $-1$, and $0$ is also a composite number?
Does it have any practical purpose?
On
A prime number by definition will have only two divisors and must be positive integers.
So by definition they start from 2 and are all positive numbers.
A composite number is also greater than 1.
On
The generalisation of prime numbers to any integral domain is the notion of irreducible elements. Let $R$ be an integral domain, an element $p$ is called irreducible if for all non unit $q \in R$, and $u \in R$, $$ p = uq \implies u \in R^\times .$$
In the integral domain $\mathbb Z$, $(-1)$ is a unit, so can't be irreducible.
The generalisation of composite numbers can be made in unique factorization domains, i.e. integral domain $R$ such that every non unit $q \in R \setminus \{0\}$ admits a factorization of the form $$ p_1 \dots p_r, $$ the $p_i$ being irreducible, unique up to permutation of terms and the multiplication by a unit. A composite element is then an element with such a factorization and $r \geq 2$.
In the UFD $\mathbb Z$, $0$ is not a non unit of $\mathbb Z \setminus \{0\}$, so can't be composite.
In a commutative ring $R$ with $1$, we say $p\in R\setminus\{0\}$ is prime if the following conditions are satisfied:
In $\Bbb Z$, $(-1)(-1) = 1$, so that $-1$ is a unit, and is therefore not prime. $0$ is not prime (using the ring-theoretic definition of prime) as we exclude it by definition, but it does behave similarly to primes: $0$ only divides $0$, and no two non-zero elements of $\Bbb Z$ will multiply to give $0$, so if $0\mid ab$, either $a = 0$ or $b = 0$, and hence $0$ will divide whichever one is $0$ (remember that by "$a$ divides $b$," we mean "there exists $k\in\Bbb Z$ such that $ak = b$," so it is fine to say "$0$ divides $0$," even though division by $0$ is undefined).
Using the above definition, we can extend the notion of primeness to not only the negative integers, but also any commutative ring! In fact, you'll find that a negative integer $p$ is prime if and only if $-p$ is prime. These primes differ by a unit $(-1)$, and in general such primes are called "associates."
A similar (but not equivalent) notion is that of irreducibility. A nonzero nonunit $r\in R$ is irreducible if it is not the product of two nonunits. In $\Bbb Z$, $p$ is prime if and only if $p$ is irreducible, but in general this is not true: in the ring $\Bbb Z[\sqrt{-5}] = \{a + b\sqrt{-5}\mid a,b\in\Bbb Z\}$, there are irreducible elements that are not prime! For example, it can be shown that $3$ is irreducible, but $3\mid 9 = (2 + \sqrt{-5})(2 - \sqrt{-5})$ and does not divide either factor!
The differences between these notions is an important aspect of number theory, and has led to many discoveries/creations, such as the notion of an "ideal," and the study of the question "in which rings $R$ does unique prime factorization hold?" One result states that for any number field (a "small" field containing the rational numbers $\Bbb Q$), in its ring of integers (the generalization of $\Bbb Z$ to fields other than $\Bbb Q$) every nonzero proper ideal will factor uniquely into the product of prime ideals, even though elements may not always factor uniquely into products of prime elements! As you may imagine, such questions are of interest to many mathematicians, and this type of generalization is very practical in mathematics - take something you know (like primeness) and extend it to apply to objects you don't understand quite as well as your original. If the proper generalization is used, one often finds surprising (and useful) analogies between the original object and the new, less-understood objects!