Desiccation cracking in expansive soils on slopes accelerates soil erosion and reduces water retention capacity, posing significant threats to agricultural sustainability. This study investigates the combined effects of organic fertilizer (OF), slow-release fertilizer (SRF), and rice straw (RS) on crack inhibition in expansive soil under wet–dry (W-D) cycles. A three-factor Box–Behnken design (BBD) was used with OF at 5–15%, SRF at 0.3–0.9 kg/m³, and RS at 0.25–0.75%, and crack development was followed through 20 W-D cycles. The quadratic response-surface model fitted to the BBD crack-ratio data was significant (R² = 0.952, adjusted R² = 0.865, p = 0.008), with significant linear effects of OF, SRF, and RS, whereas the two-factor interaction terms were not statistically significant as a group (p = 0.701). The lowest observed crack ratio in the BBD matrix was 1.7% at 10% OF + 0.9 kg/m³ and SRF + 0.75% RS; this is therefore reported as an observed low-response combination rather than evidence of a statistically confirmed SRF–RS interaction. An LSTM model was retained as a complementary dynamic predictor and yielded R² = 0.927, MAE = 0.325, and RMSE = 0.478 for the evaluated dataset; however, divergence between training and validation losses indicated overfitting, so external generalizability remains to be established. The results support agricultural-waste-based amendment as a promising crack-control strategy while emphasizing the need for independent replication and field validation.