The effectiveness of robotic exoskeletons for post-stroke gait rehabilitation might be limited as the control parameters of these devices do not adapt to key biomechanical descriptors. The main contribution of this study is to examine post-stroke gait with the aim of finding relationships between exoskeleton control parameters and a comprehensive set of biomechanical metrics. Five stroke survivors walked with the assistance of a wearable ankle exoskeleton (ABLE-S) using different levels of plantarflexion (PF) and dorsiflexion (DF) peak torque, as well as different timings of PF peak torque. We found that DF peak magnitude had significant negative relations with the temporal symmetry index (p = 0.033) and the paretic foot absolute angle at heel strike (p = 0.019). Changes in the applied PF assistance parameters were significantly correlated with a high variety of temporal and spatial parameters, e.g., walking speed (p = 0.009), stride length (p = 0.011), non-paretic step length (p = 0.024), foot clearance (p = 0.003) and hip hiking (p = 0.038), and the muscle activation for the non-paretic side, e.g., Tibialis Anterior (p = 0.049) and Gastrocnemius Medialis (p = 0.049). Based on our results, we propose a set of control laws for adapting the assistance of ankle exoskeletons that will be evaluated in future work.