Non-contact, fluid-mediated manipulation of objects using underactuated soft robots or structures presents unique control challenges, demanding a nuanced understanding of fluid dynamics and innovative control methods. In this study, we investigate methods for efficiently manipulating objects in water, exploiting the fluid structure interactions of soft underactuated soft tentacles. To leverage fluid interactions, we identify behavioral primitives that enable object manipulation through different modalities of vortex generation, and demonstrate this through PIV imaging. This reduced order control space can be used to aid both teleoperation-based manipulation, and also to develop a Finite State Machine (FSM) based controller, which offers an efficient means to exploit fluids for manipulation. This lays the ground for more data intensive approaches for learning controllers for this complex control challenge.