Working memory (WM) has a limited capacity; however, this limitation can be mitigated by selecting individual items from the set currently held in WM for prioritization. The selection mechanism underlying this prioritization ability is referred to as the focus of attention (FOA) in WM. Although impressive progress has been achieved in recent years, a fundamental question remains unclear: Do perception and WM share one FOA? In the current study, we investigated the hypothesis that only a perceptual task tapping object-based attention can divert the FOA in WM. We adopted a retro-cue WM paradigm and inserted a perceptual task after the offset of the cue. Critically, we manipulated the type of attention (object-based attention in Experiments 1-3, feature-based attention in Experiment 4, and spatial attention in Experiment 5) consumed by the perceptual task. We found that participants were able to prioritize a retro-cued representation in WM, and the retro-cue benefit on memory accuracy was intact regardless of the perceptual task. Critically, the retro-cue benefit on the response time of WM task was significantly reduced only after an object-based attention perceptual task (Experiments 1, 2, 3a, and 3b), while remaining constant after a feature-based attention (Experiment 4) or spatial attention (Experiment 5) perceptual task. These results suggest that WM and perception share an object-based FOA, and an object-based attention perceptual task can divert the FOA in WM. Meanwhile, the current study further confirms that sustained attention is not necessary for selective maintenance in WM.