Let $\mathscr{F}$ be some $\mathscr{O}_X$-module. To say what Georges has expertly said in a different way (I hope):
(1) To me, $\mathscr{F} \otimes k(p)$ usually means "pull back $\mathscr{F}$ to the point $\operatorname{Spec} k(p)$". Since that's just a point, we can identify the sheaf with the module: it's $\mathscr{F}_p \otimes_{\mathscr{O}_{X, p}} k(p) = \mathscr{F}_p/\mathfrak{m}_p\mathscr{F}_p$.
(2) Working over some field $k$ (it could be a ring; I don't think it matters), the constant sheaf $\underline{k}$ with values in $k$ is a subsheaf of rings of $\mathscr{O}_X$; and the constant sheaf with values in $H^0(X, \mathscr{F})$ is a sheaf of modules over that, so we can tensor over $\underline{k}$ with $\mathscr{O}_X$. I think this is usually done in order to talk about global generation, ie, whether the multiplication map $H^0(X, \mathscr{F}) \otimes \mathscr{O}_X \to \mathscr{F}$ is surjective.