Children with cerebral palsy must perform daily exercise which is a tedious and energy consuming task. Exergames can make this routine more engaging, which can increase the compliance of the patient. This research explores the feasibility of an exergame device called the Squid Monster. The device is the result of a research through design process, and it is designed to be played on smartphones in the home environment. It operates on the smartphone's integrated sensors and two external squeeze sensors, making it accessible and cost-effective. We conceptualize how the design can be supported using a machine learning adaptive difficulty system, aiming to increase flow and therapeutic adherence of the device. Ultimately, guidelines are provided to designers for future work in this field.