Approximately one-third of fresh fruits are lost in the supply chain, with half the losses occurring during in-field harvesting and in-plant grading. The complexity of supply chain, waste and economic losses could be reduced through integrated non-destructive harvesting and in-field quality grading. Here we developed a self-decoupled tactile sensor for robotic soft gripper, embedding three-axis self-decoupled triboelectric soft tactile sensors across fingers. During single harvesting operation, the strategy can measure the gripping force, fruit firmness, weight by tactile sensors. Hence, integrated non-destructive harvesting and in-field quality grading can be achieved. Tactile sensors are bioinspired by the deformation mechanics of octopus suckers under normal and tangential forces. It features dual-mode TENG, integrating vertical contact-separation and horizontal sliding modes. This structure can independently respond to normal and tangential forces. It will generate a superimposed triboelectric signal with three spatially decoupled components. Consequently, the three-axis force can be easily decoupled with three triboelectric signals independently, achieving high-precision measurement within the 0-20 N range. Finally, we demonstrate the practical efficacy of the sensing strategy through in-field experiments. During grasping, fruit firmness is evaluated from the normal force response over time. Upon fruit detachment, tactile sensors monitor both normal and tangential forces in real-time to prevent damage. Finally, with the gripper reoriented palm-down, the fruit weight is non-destructively calculated via static equilibrium between the tangential force and the fruit's gravity. We develop a robotic sensing system for integrated, non-destructive fruit harvesting and in-field grading, thereby streamlining the supply chain, reducing waste, and enhancing food safety.