To do this systematically, with the rules of Hoare logic, proceed as follows.
First, use the assignment axiom to generate a valid Hoare triple of the form $\{\phi\}n \mathbin{:=}a-b\{n<a \land a+b\geq 0\}$.
Then use the consequence rule to derive the Hoare triple you're after. Besides the Hoare triple we just found, this instance of the consequence rule will have a premise $a>7 \land b\geq 0 \Rightarrow \phi$. You need to prove that it is valid, too, but that's a maths task independent of Hoare logic. As Yuval observed in his informal explanation, you will fail to prove this.