Constructing multi-metal centers on carbon-based substrates is a promising strategy to enhance C-N coupling for efficient urea synthesis, while the underlying design principles, particularly how metal-metal and metal-substrate interactions govern reactant activation and reaction pathways, remain intangible. To address this gap, we developed a frontier orbital interaction-guided C-N coupling selectivity map based on the p-d asymmetric dual-atom models (DACs) through synergistic integration of DFT calculations and machine learning classification. Specifically, efficient NOx reduction was found to require a narrow energy gap (ΔE1 < 3.38 eV) between the HOMO of p-block metals and the LUMO of the d@substrate (where d-block atoms are treated as integrated with substrates for simplification). In contrast, selective urea synthesis necessitates a larger energy gap (ΔE2 > 1.39 eV) between the LUMO of p-block metals and the HOMO of the d@substrate, signifying weaker p-d interactions. Moreover, such asymmetric dual-atom structure enables a tunable bandgap while simultaneously optimizing visible-light absorption range. As a result, the AlPd@PCN and GaPt@PCN systems stand out as exceptional candidates, exhibiting fully thermodynamically favorable energy profiles throughout the photocatalytic cycle. These insights not only extend frontier orbital theory to DACs systems but also establish a robust, generalizable framework for designing high-performance dual-atom urea synthesis catalysts.