The development of assistive lower-limb exoskeletons can be time-consuming. Testing prototype medical devices on vulnerable populations such as children also has safety concerns. Mechanical phantoms replicating the lower-limb kinematics provide an alternative for the fast validation and iteration of exoskeletons. However, most phantoms treat the limbs as rigid bodies and fail to capture soft tissue deformation at the human/exoskeleton interface. Human soft tissue can absorb and dissipate energy when compressed, leading to a mismatch between simulated and human exoskeleton testing outcomes. We have developed a methodology for quickly testing and validating the performance of knee exoskeletons using a mechanical phantom capable of emulating knee kinematics soft-tissue deformation of the lower-limb. Our phantom consisted of 3D-printed bones surrounded by ballistic gel. A motorized hexapod moved the knee to follow a walking trajectory. A custom inverse dynamics model estimated the knee assistance moment from marker and load cell data. We applied this methodology to quantify the effects of soft-tissue deformation on exoskeleton assistance by loading the phantom knee with a torsional spring exoskeleton interfacing and bypassing the ballistic gel. We found that including soft-tissue deformation led to a lower knee assistance moment and stiffness. Some but not all of this difference could be explained by the deflection of the exoskeleton relative to the knee angle, suggesting energy absorption within soft tissue. The direct measurements of exoskeleton assistance provide insight into increasing the assistive moment transmission efficacy. The phantom provided a relatively accurate framework for knee exoskeleton testing, aiding future exoskeleton design.