Converging evidence suggests that traditional domain-general working memory (WM) training does not have reliable far-transfer effects, but produces reliable, modest near-transfer effects on structurally similar untrained tasks. Given the critical role of WM in academic development, WM training that incorporates task-specific features may maximize training effects on academic outcomes. In this theory paper, we discuss the training to emphasize the domain-specific function of WM highlighted by recent WM models. That is, WM should be better attuned to the materials being learned through enhancing strategies of linking together WM with the long-term memory knowledge, rather than only the enhancement of a "domain-general" attentional control overall. We provided two example training routes that emphasize explicit instruction and practice on WM-academic tasks (i.e., academic tasks that can be performed using a WM training paradigm) and task-linking strategies (i.e., strategies that can be used in both academic tasks and WM tasks to improve performance efficiency). We also review recent relevant intervention studies that are in line with this approach and report promising effects on academic outcomes. Implications for future studies are also discussed.