The robotic ultrasound system has the potential to improve the conventional practice of diagnosing. Because of the adequate degrees of freedom embedded in a small footprint, the parallel mechanism-based ultrasound robot has attracted attention in the field. However, the analysis of its configuration, design parameters, and workspace is limited. To solve this issue and further promote the potential clinical translation, this paper proposes a task-driven, two-stage mechanism optimization method using the effective regular workspace and the local condition index to determine the parameters for the demanding clinic workspace of a parallel mechanism-based ultrasound robot. The design and implementation method of the robot are then introduced, along with the justification of parameter selection. To analyze the performance, an optical tracking-based experiment and a phantom-based human-robot comparison study were performed. The results show that the workspace meets the required clinical needs, and despite its small footprint, the mechanism could have a reasonable workspace. The kinematic error was found to be 0.2 mm and 0.3°. Based on the above results and the quantitative analysis of the ultrasound images acquired manually and robotically, it was concluded that the robot can effectively deliver the demand function and would be a promising tool for further deployment.