Consistent orientation of surface normals is a global problem in computational geometry. Orientation is intrinsically global: local geometry determines a normal line, but not which of its two directions is outward. We study the orientation problem at the continuum level for a smooth closed surface $\Gamma=\partial\Omega$. Starting from the Newtonian single-layer potential, we consider its averaged boundary divergence and curl traces, denoted by $B$ and $C$. The classical scalar Gauss constraint
$B\varphi=1/2$ alone does not determine a free vector density. We show that adding the curl condition $C\varphi=0$ makes the system complete. More precisely, for every $s\in\R$,
\[
\|\varphi\|_{H^s(\Gamma;\R^3)}
\le C_{\Gamma,s}
\bigl(\|B\varphi\|_{H^s(\Gamma)}^2+
\|C\varphi\|_{H^s(\Gamma;\R^3)}^2\bigr)^{1/2}.
\]
$P=(B,C)$ is therefore injective with closed range, and $P\varphi=(1/2,0)$ has the unique solution $\varphi=n_\Omega$. The
Clifford Cauchy involution $T^2=I/4$ is classical. The new result proved here is the same-order lower bound for $P=(B,C)$; its problem-specific proof step is the factorization $TM=JP$, which identifies $B$ and $C$ with the scalar and bivector parts of the same Cauchy singular integral. This factorization also yields a compatibility condition for general boundary data. If the samples are distributed according to a positive density $\dd\mu=\rho\,\dd S$, the weighted system recovers $q=n_\Omega/\rho$ uniquely, so normalization removes the unknown sampling density and returns the outward orientation. Regularized point-cloud experiments illustrate the continuum result.