Ok, this question makes more sense to me than your other (earlier) question.
$G$ doesn't act on local sections. $G$ isn't guaranteed to preserve open sets, (i.e. $gU\
e U$ necessarily) so how could it possibly act on $\
ewcommand\calO{\mathcal{O}}\calO_X(U)$ for all open $U$?
Instead, if $U$ is $G$-invariant, $G$ descends to an action on $\calO_X(U)$ in the obvious way. Note, however that $G$ acts by ring homomorphisms not module morphisms.
In general, the action of $G$ on $X$ will not induce an action of $G$ on a particular $\calO_X$-module. For example, if we take a sky scraper sheaf over some point with nontrivial orbit, then clearly $G$ will act by sending the sky scraper sheaf to a sky scraper sheaf with a different base point.