This paper presents a covariance-aware KiDS-1000 PneE weak-lensing projection test in the ECSM framework. The analysis compares null, constant-geometry, smooth geometry, frozen boundary-control, external-geometry response, and domain-boundary-gradient response models. The V6 ECSM domain-gradient projection gives the best full-data performance, with chi2 = 141.68 compared with chi2 = 209.51 for the smooth geometry control and chi2 = 1498.48 for the null model. It also transfers across held-out source-bin and lens-bin validation, improving over the smooth control by Delta chi2 = 79.62 and Delta chi2 = 186.01 respectively. The result is presented as a projection-bridge validation rather than a final first-principles baryonic-mass prediction, because the tested KiDS bandpower file does not include physical lens baryonic masses, luminosities, sizes, or scale radii.