The seems a really weird way of putting it but it would work.
If for all $x,y \in T$ we always have $xy^{-1}\in T$ then
For any $x,x\in T$ then $xx^{-1} = e\in T$ and therefore for any $x\in T$ then $e\cdot x^{-1} =x^{-1} \in T$.
And finally if $x,y\in T$ then $y^{-1} \in T$ (as we showed above) so $x(y^{-1})^{-1} = xy\in T$.
That's all we need.
We've proven $\cdot$ is closed (but IMO in a really round about way). Associativity is inherited. The identity element of $\mathbb C$ is the identity element of $T$ and we've proven (very round about way) that $e=1\in T$. And we have proven for every $x\in T$ that $x^{-1}\in T$.
That's all that is needed to show its a group.
.....
But in my opinion this "let's do it by proving only one thing" is cutting your nose of to spite your face. And I'm not sure how you'd prove $xy^{-1}\in T$ without first showing $y^{-1}\in T$ and frankly proving the TWO items 1) $|1|=1$ so $1 \in T$ and 2) if $|x|=1$ then $|x^{-1}| = 1$ so for all $x \in T$ we know $x^{-1}\in T$, would be a proof that is easier, more direct, and much more readable.