A critical problem in robotic manufacturing is that the task of teaching robotics is rather time-consuming. This has become a serious problem in the present age of cost reduction. Collaboration with a company in the field has revealed that the root cause of this problem is that there is not a common knowledge base in this domain, which can serve as shared and reused knowledge. In robotic manufacturing, the skills and experiences of skilled workers are a form of tacit knowledge that is difficult to be acquired and transferred to other workers and robots. This paper proposes a knowledge-based system for sharing and reusing tacit knowledge in the robotic assembly domain. In this system, a modified EBL (Explanation-based Learning) method is proposed to generalize tacit knowledge from specific robotic programs made by skilled workers. A newly operational criterion is proposed for the generalized tacit knowledge, which demands that it should be expressed understandably by human workers and be reusable by robots to generate programs automatically.