Hand interaction has gained popularity as an alternative to controller-based interactions in commercial mixed reality headsets. However, current interaction paradigms remain heavily influenced by desktop computers and smart devices, often relying on the WIMP (Windows, Icons, Menus, Pointer) paradigm. To better leverage the user’s physical context, this paper presents an empirical gesture design study for a cooking assistant mixed reality application, where user gestures dynamically adapt to the object held in their hand, enabling more natural and context-sensitive interactions. We propose the usage of a co-design study to complement standard gesture elicitation techniques and explore how the shape, weight, and perceived dangerousness of cooking tools influence gesture design.