Intensive pesticide application sustains global agriculture but poses poorly characterized risks to complex soil ecosystems. Here, we quantitatively evaluated pesticide residues and utilized high-resolution environmental DNA (eDNA) metagenomics to decode multi-trophic community responses across a typical major grain-producing region located in China. Among 39 targeted pesticides, 26 were detected with total concentrations ranging from 27.9 to 478.8 ng/g. While herbicides and fungicides dominated the residual mass, insecticides posed the most severe ecological threat. Notably, the neonicotinoid imidacloprid exhibited high-risk levels (RQ = 1.78 ± 1.49) at >61.1% of the sampling sites. eDNA profiling and Procrustes analyses revealed a clear trophic-dependent sensitivity gradient (p < 0.01). Lower-trophic microbial communities were significantly altered in composition; pesticide stress was strongly associated with profound non-target suppression on keystone plant-beneficial bacteria (e.g., Nocardioides). Concurrently, the fungal eDNA profiles indicated that the soil mycobiome harbored an alarming 34.7% of potential phytopathogenic fungi (e.g., Aspergillus and Colletotrichum), intrinsically driving the massive fungicide reliance. In contrast, higher-trophic soil metazoa (Rotifera, 40.4%) and weed communities (e.g., Digitaria sanguinalis) exhibited significant spatial stability, reflecting robust environmental buffering and herbicide-driven ecological escapes. Furthermore, co-occurrence networks decoupled target from non-target toxicities, uniquely revealing that persistent herbicide metabolites (desethylatrazine) induce prolonged legacy toxicities on specific soil fauna. Collectively, this study unveils the deep, cross-kingdom ecological disruptions caused by current pesticide regimes, underscoring the urgency of integrating eDNA biomonitoring to guide precision pest management and safeguard soil health in vital agricultural hubs.