For (2), I suggest writing $a$ as the Turing join of its even and odd parts, call them $a_0$ and $a_1$.  Use $a_0$ to compute a perfect subtree $T$ of $2^\omega$ such that every infinite path in $T$ is mutually Cohen generic with $a_1$.  When considered as a Sacks condition in $L[a]$, $T$ forces that $a$ is not an element of $L[b]$.