We discuss steps towards a formalisation of the principles of an agentive naïve proto-physics, designed to match a level of abstraction that reflects the pre-linguistic conceptualisations and elementary notions of agency, as they develop during early human cognitive development. To this end, we present an agentive extension of the multi-dimensional image schema logic ISL based on variants of STIT theory, thus replacing the temporal dimension of ISL with an action-agnostic theory of agency. To begin grasping the notion of ‘animate agent’, we apply the newly defined logic to model the image schematic notion of ‘self movement’ as a means to distinguish the agentive capabilities of a mouse from those of a ball. Finally, we outline the prospects for employing the theory in cognitive robotics.