This TestBed address one of the most stubborn unsolved bottlenecks in electronics manufacturing: dexterous handling and insertion of flexible cables and connectors during server-tray assembly. TestBed enables training of AI policies in simulation — with an open-source simulator as the official evaluation environment and alternative simulators permitted for training. The reference hardware stack comprises a 6-DoF collaborative robot arm, a parallel-jaw gripper, a wrist force-torque sensor, and three wrist-mounted RGB cameras streaming at 20 fps, all interfaced through standard ROS 2 (Kilted Kaiju) topics, services and actions. The TestBed provides high-fidelity SDFormat scene descriptions, URDF/SDF robot models, a baseline controller and hardware abstraction layer, a ROS 2 lifecycle node template that dynamically loads Python policies, an automated scoring engine, example policies, container images, and a reproducible development environment, supporting Industry 5.0 research on sim-to-real, human-centric dexterous robotic manipulation.