Understanding the evolutionary characteristics of regional droughts is essential for ecological and socioeconomic sustainability, especially in water-scarce regions. The propagation from meteorological drought (MD) to agricultural drought (AD) in the Loess Plateau (LP), shaped by climate change and large-scale vegetation restoration, has not yet been fully elucidated. This study investigates the dynamic propagation from MD to AD in the LP, employing cumulative precipitation anomaly and the standardized soil moisture index to characterize MD and AD, respectively. A mutation detection algorithm was applied to divide the study period into pre- and postmutation phases, allowing for a systematic examination of changes in drought propagation characteristics and their controlling factors. The results indicate that, following the mutation, the mean drought propagation time from MD to AD across the LP decreased from 5.0 to 4.5 months. With increasing severity of MD, the drought propagation risk exhibits a nonlinear rise, and this risk further intensifies after the mutation. The required to trigger mild AD declined from 62.2 to 45.5 mm, suggesting reduced resilience of the agricultural system to MD. Dynamic attribution analysis further revealed that the normalized difference vegetation index was the dominant driver of both the shortened propagation time and the reduced thresholds, suggesting that vegetation restoration has, to some extent, exacerbated the propagation from MD to AD. These results not only contribute to a deeper understanding of the dynamic propagation mechanism from MD to AD under changing environmental conditions in the LP but also offer scientific guidance for developing targeted drought-resistance strategies and mitigating the risk of AD.