This paper presents a variable damping control strategy to improve trade-off between agility/performance and stability in the control of the ankle exoskeleton robot. Depending on the user's intent of movement, the proposed variable damping controller determines the robotic ankle damping from negative to positive damping values. The range of damping values is determined by incorporating the knowledge of human ankle damping in order to always secure stability of the ankle joint of the coupled human-robot system. To evaluate the effectiveness of the proposed controller, we performed a set of human experiments with three different robotic damping conditions: fixed positive damping, fixed negative damping, and variable damping. Comparison of the two fixed damping conditions confirmed that there exists a clear trade-off between ankle agility and stability. Further, analysis of the variable damping condition demonstrated that humans could get benefits of not only positive damping to stabilize the ankle but also negative damping to enhance the agility of ankle movement as necessary during dynamic ankle movement. On average, the variable damping condition improved the agility of ankle movement by 76% and stability by 37% compared to the constant positive damping condition and the constant negative damping condition, respectively. Outcomes of this study would allow us to design a robotic controller that significantly improves agility/performance of the human-robot system without compromising its coupled stability.