Advancements and the adoption of productive human-robot collaborations have im pacted every aspect of modern manufacturing. Although automation has been increasingly utilized on the shop floor for decades, a significant portion of manufacturing operations remains manual primarily due to the dominating ability of human operators to perform these operations well, compared to robots. These manual or semi-automated operations involve timely complex manipulations and reasoning, and depend highly on human labor skills, intelligence, and expertise. To prepare the workforce of the future, manufacturers are increasingly turning to augmented reality (AR) to prepare, train, guide, and refine the skills of their less-experienced workers by allowing skilled workers to collaboratively guide and mentor less-experienced workers, collaboratively and on the fly. In this paper, we propose a new skill- and- knowledge- -sharing model in manufacturing systems with a hub for collaborative intelligence (HUB-CI) that supports an effective learning protocol. The learning curves of operators are modeled as a function of time and the complexity of skills and knowledge, and analyzed for the learning quality. A case- study of assembly kitting workflow is used to conduct the experimental simulations and analyses. Several insights are drawn from the preliminary simulation analyses to inform the design of future AR technologies for skill and knowledge sharing among manufacturing workers, working collaboratively with robots on complex manipulation and reasoning tasks.