Shahriar Shahriari proved in On normal subgroups of capable groups, Arch. Math. (Basel) 48 (1987) no.3 193-198, MR 0880078 (88e:20026), among other things, that if a finite group $G$ contains a normal subgroup $H$ which is either generalized quaternion of order $2^n$, $n\gt 2$, or semidihedral of order $2^n$, $n\gt 3$, then $G$ is not capable. This immediately shows that $Q_8\times Q_8$ cannot be capable (and more generally, gives you lots of examples of groups $H$ such that $H\times H$ is not capable; note that if $H$ is capable, then so is $H\times H$, since any witness $K$ to the capability of $H$ yields $K\times K$ as a witness for the capability of $H\times H$).
Shahriari also proved that if $G$ is finite nilpotent and contains a normal subgroup $H$ which is an extraspecial $p$-groups of order $p^3$ and exponent $p$, $p$ odd, then $G$ is not capable.
Finally, Shahriari proves the following:
Theorem (Prop. 3.2 in the above-cited paper) Let $G$ be a finite group. If $G = QC_Q(G)$ for some $Q\leq G$, and $1\neq M\subseteq Z^*(Q)\cap[Q,Q]$, then $M\subseteq Z^*(G)$; in particular, $G$ is not capable.
Here, $Z^*(Q)$ is the epicenter of $Q$ (similar with $G$). The epicenter is the obstruction to capability: it is the smallest central subgroup of $G$ such that $G/Z^*(G)$ is capable.