Stunting in children under five remains a critical public health problem in Indonesia, particularly in rural areas where access to optimal complementary feeding practices and health literacy are limited. The study was conducted in Bone Regency, Indonesia. Aim: This study aimed to develop and validate a predictive model to identify infants aged 6–12 months at risk of stunting (LAZ < -2) using maternal behavioral determinants and feeding indicators. Methods: A cross-sectional design was used to combine online questionnaire responses from 401 mothers with anthropometric data. The model incorporated constructs from the Health Belief Model (HBM)—including perceived severity, benefits, barriers, self-efficacy, and cues to action—in feeding based on WHO/UNICEF criteria (MDD, MMF, MAD). Structural equation modeling (PLS-SEM) was applied to test direct and indirect pathways to risk classification. Findings: The study showed that feeding practices were the strongest direct predictor of stunting risk (β = 0.537, p < 0.001). Mediation analysis confirmed that barriers indirectly increased risk through their negative impact on feeding behavior. The model demonstrated acceptable fit (SRMR = 0.061) and explained substantial variance in the primary outcomes (R² = 0.825 for Intention, 0.526 for Feeding Practices, and 0.542 for Risk). Conclusion: These findings suggest that targeting maternal beliefs and behavioral intentions can increase complementary feeding and reduce the risk of stunting. The intervention focused on reducing perceived barriers, strengthening self-efficacy, and ensuring timely cues to action. Community health centers can play a critical role in operationalizing this strategy at scale.