During the battle against COVID-19, sudden-onset disasters keep happening. These two consecutive events are different in nature and sometimes ask for counter-productive policy interventions: where the epidemic can be contained by ensuring social distancing practices, reducing the number of contacts, and offering hygienic conditions, the response to a natural hazard requires evacuation, which often leads to overcrowded shelters with a questionable hygiene. An exploratory agent-based model is constructed where the dynamics of these socio-technical systems are integrated to find high-level emergent behaviour over time during the response phase of a sudden-onset disaster. The study describes a stylistic agent-based model that combines three socio-technical systems, their interactions and system behaviour to find general trends and interdependencies. The three sub systems comprise of a COVID-19 component, based on the SEIR approach, a livelihood component, which mimics a microeconomic marketing mechanism and accounts for among others difference in occupation, and a hazard component, which is responsible for the evacuation process. The following research question is answered: What robust policy interventions can be identified that balance livelihood of rural communities and their exposure to COVID-19 during the response to a sudden-onset disaster in developing countries? The goal is to find robust policy interventions that can be executed by either local governments or humanitarian organizations and are suitable for poor rural communities in developing countries. Four policy interventions were compared: cash transfers, an awareness campaign, increasing the number of shelters, and varying the moment of imposing a lockdown. With the use of Exploratory Modelling and Analysis, an experimental design was constructed to identify uncertainties and model behaviour. Results show that using direct and unconditional cash transfers in combination with increased awareness have most beneficial effects for the average livelihood and also positively influence the COVID-19 trajectory. Increasing awareness shows best results for containing COVID-19, but is dependent on context factors such as regular testing and implementation before the the virus spirals out of control. Awareness also negatively influences the average livelihood. This research is a first step in modelling compound risk of COVID-19 and sudden-onset disasters. The modular way of building makes this research useful not only for searching trade-offs in developing countries, but can be used in other contexts as well.