<p>Precise risk identification of heavy metal(loid)s in agricultural soils is pivotal for food safety, yet conventional generic standards often obscure localized hazards and misdiagnose key drivers. Herein, we developed an integrated diagnostic framework that combines local geochemical baselines (GBLs), a multi-index pollution assessment, bioavailability-informed ecological risk evaluation, and a source-oriented probabilistic human health risk assessment using positive matrix factorization coupled with Monte Carlo simulations. This framework was applied to a mining-impacted agro-ecosystem in the Huicheng Basin, China. Results revealed substantial accumulation of As, Cd, Cr, Cu, Ni, Pb, and Zn, with mean Cd and Zn concentrations exceeding local GBLs by 3.6-fold and 2.8-fold, respectively. Ecological risks were primarily driven by the high bioavailability of Cd and Pb rather than total concentrations. Source apportionment identified three major contributors: mining activities (47.60%), agricultural inputs (34.97%), and natural origins (17.43%). Notably, a source-risk decoupling was observed: while mining dominated total metal mass, agricultural inputs accounted for the highest carcinogenic risk to children (48%), mainly due to As and Ni. These findings highlight that conventional mass-loading reduction strategies may overlook key high-toxicity, high-bioavailability sources, emphasizing the need for targeted management that addresses both ecological and human health risks in mining-affected agricultural soils.</p>