This paper relates research in progress illustrating social-cognitive robotics in manufacturing, where a social robot (QT) materialises a natural interface between the manufacturing environment and the human operator through its digital counter part, a Human Digital Twin (HDT). One future goal is that QT proactively and cognitively assesses certain mistakes that can occur during a product assembly task in order to mitigate risks that can harm the user or damage products. Here, we present a literature review on social-cognitive robotics, our approach involving HDT and the QT with LLM-based cognitive functions, and initial results of an exploratory user study with a Lego assembly.