Living systems adopt diversity of shapes. These morphological changes, which range from cell to organismal scales, commonly rely on intrinsic contractile stresses generated by myosin in the cell cytoskeleton, an adaptive active contractile filamentous material, whose molecular constituents, convert chemical energy to produce mechanical work. How these intrinsically active stresses arise in complex 3D shapes and how shape deformation is controlled remains poorly understood. Here, we demonstrate that, initially homogenous not-prepatterned elastic actomyosin networks spontaneously self-organize in a family of 3D shapes through distinct dynamical patterns. Shape selection is encoded in system size to thickness aspect ratio, indicating shaping scalability. The final configurations show surprisingly simple scaling dependence on system dimensions. Altogether, cytoskeletal networks form a class of adaptive active materials, autonomously designing their own shape in response to system geometry, without needing specific pre-programming. This simplicity present huge advantage for developing bio-soft robots with desired target shapes.