Abstract-Robotic systems are increasingly used in healthcare tasks such as medicine delivery, where both kinematic performance and efficiency are critical. This paper presents a task-based, manipulability-aware motion planning approach for a 7-DOF robotic arm. A finite state machine is used to manage task phases—rest, approach, delivery, and return—while cosine-based joint interpolation ensures smooth motion. Manipulability is evaluated using the Jacobian, and a gradient-based optimization term is introduced to enhance end-effector dexterity in real time. The approach leverages kinematic redundancy to maintain high manipulability throughout task execution. Simulation results in Webots demonstrate improved average and peak manipulability without compromising task accuracy. The robot achieves precise positioning, smooth motion, and reliable interaction, validating the effectiveness of the proposed method for healthcare applications.