We study the transition probability, say $p_A^n(x,y)$, of a one-dimensional random walk on the integer lattice killed when entering into a non-empty finite set $A$. The random walk is assumed to be irreducible and have zero mean and a finite variance $σ^2$. We derive the asymptotic form of $p_A^n(x, y)$ for large $n$ valid uniformly in the regime characterized by the conditions $|x|\vee |y| =O(\sqrt n)$ and $|x|\wedge |y|= o(\sqrt n)$, in which $p^A_t({\bf x},{\bf y})$ behaves for large $n$ like $[g_A^{+}(x)\hat g_{A}^{\,+}(y) + g_A^-(x)\hat g_{A}^{\,-}(y)] (σ^{2}/2n) p^n(y-x)$. Here $p^n(y-x)$ is the transition kernel of the random walk (without killing); $g^\pm_A$ are the Green functions for the "exterior" of $A$ with "pole at $\pm \infty$" normalized so that $g^\pm_A(x) \sim 2|x|/σ^2$ as $x \to \pm\infty$; and $\hat g_A^{\, \pm}$ are the corresponding Green functions for the time-reversed walk.