Collaboration between robots and human workers constitutes an important part of the shift towards human-centric production processes. While collaborative robots have increased flexibility and enabled compact production environments, real-time intent recognition and joint actions in human-robot collaboration (HRC) continue to pose challenges, particularly in dynamic target selection and coordination. Conventional HRC systems rely on manual inputs or predefined gestures. This study explores the feasibility of gaze-based interaction as a natural and intuitive control mechanism for close-proximity HRC in dynamic industrial settings. The approach is demonstrated in the form of a prototype workstation for wire harness assembly on a car bumper, where the human and the robot collaborate in real time using eye gaze-based dynamic target selection and tight action coordination. The human eye gaze was recorded in real-time using eye-tracking glasses, and the forward-facing scene camera of the glasses matched it to upcoming assembly targets for the robot. The system was evaluated based on detection accuracy and response time. Results indicate that the human eye gaze can constitute a powerful cue for robotic target selection and action coordination in HRC-based assembly. The YOLOv8-based object detection model achieved average target recognition confidence above 60%, despite sensitivity to lighting variations and dataset composition. These findings support the viability of gaze-based robotic control in industrial assembly, demonstrating its potential to enhance operator ergonomics, streamline task execution, and improve overall system responsiveness in dynamic work environments.