A pathway-based diagnostic framework is proposed to bridge perturbed parameter ensemble (PPE) sensitivity analysis and process-oriented moist static energy (MSE) budget diagnostics for evaluating Madden–Julian Oscillation (MJO) simulation in general circulation models. The framework connects statistical parameter sensitivities to the physical processes that shape MJO behavior and provides a mapping between influential parameter perturbations and process-level pathways. Applied to the Community Atmosphere Model version 6 (CAM6) using a 128-member PPE, the framework identifies three dominant parameters: the fractional entrainment rate (<em>dmpdz</em>) and the convective adjustment timescale (<em>tau</em>) in the Zhang-McFarlane deep convection scheme, and the stratiform ice fall speed (<em>ai</em>) in the Morrison-Gettelman version 2 microphysics scheme. Targeted experiments show that improved MJO skill is associated with two well-supported pathways. The first pathway involves an increase in <em>dmpdz</em>, which suppresses premature deep convection and restores the equatorial low-level moisture distribution, enabling realistic meridional MSE advection that drives MJO eastward propagation. The second pathway involves a reduction in <em>ai</em>, which prolongs the residence time of upper-tropospheric ice and enhances high cloud cover along the MJO propagation pathway, strengthening the developmental-phase longwave cloud–radiation feedback that supports MJO maintenance. The role of <em>tau</em> is also important, but its specific pathway cannot be fully isolated within the present experimental design. The improved MJO simulation does not degrade the mean precipitation metrics examined here, indicating that the targeted parameter changes do not come at the expense of the simulated mean rainfall pattern in CAM6. This framework provides a process-based strategy for model evaluation and parameter calibration in general circulation models with comparable convection and cloud microphysics schemes.