The objective of this article is to present a comprehensive task analysis methodology that can provide guidelines for the design of dexterous robotic grippers that are versatile enough to perform various tasks, yet simple to manufacture. This methodology combines a human-centered gestures analysis and an object-centered grasp stability analysis. The former relies on a careful examination of a human operator’s hands gestures while performing a specific process, providing designers with tools that help specify the number of fingers, the number of degrees-of-freedom, and the placement of tactile sensors. The latter exploits a grasp quality metric to compute the efforts required to handle the involved objects, providing guidelines for the specification of the actuation system. Using observations of operators at work as a source of inspiration allows guarantying the ability to perform the considered tasks (with guaranteed stability, thanks to the grasp analysis), contrary to technologically driven optimization methodologies, which often sacrifice manipulation capabilities for the sake of simplicity. Yet, our task-oriented approach allows focusing on certain tasks, hence simpler solutions than bio-mimetic designs that try to fully mimic the human hand. In other words, the methodology introduced in this article intends to help specify multi-fingered architectures able to maintain a high degree of dexterity with a reduced kinematic complexity, favoring the best possible compromise between grasp capabilities and design complexity. This approach is exemplified by defining technical specifications for the design of a multi-fingered robotic gripper intended to perform the tasks involved in a sterility testing process.