The title of the article clearly stated the presented product and its application towards prosthesis. The abstract further elaborated on the performance metrics and its application demonstration. By combining resistive and capacitive sensing principles, the authors utilized their strength as bending sensors and pressure sensors respectively. I recognize a few prestigious review article and previous work of impact in the reference section.

The authors started the introduction by discussing the significance of touch sensing for human beings. Artificial reconstruction of tactile sensing are mentioned as robotic and prosthetic applications. The authors pointed out the unmet demand of wearable e-skin instead of existing discrete sensing elements. Two shortcomings were discussed: undesired wrinkle and fail when placed over articulations, as well as systematic characterization in real-life conditions. Since the human skin wrinkles when bent over joints, I don't see the first point as a challenge for tactile sensors, unless it is prone to fail under compressive load. I am more interested in how the second challenge is addressed by applying dynamic composite loading states to a finger/hand glove with distributed and different sensing elements.

The authors presented their work at the end of the introduction as a finger glove mounted with distributed pressure sensing elements and finger movement sensing. The incorporation of metallic shielding for capacitive sensors would greatly increase their utility in real-life environment. The sensing range of the pressure sensors are relevant, however the sensitivity value by itself holds little merit without additional measures like resolution, non-linearity, repeatability, drift, etc. Gold strain gauge with liquid metal lead wires is an interesting fabrication approach.

Overall, the combination of the two sensing elements made use of their strength, and different modalities of tactile information were found rich and useful. By providing rich information on how the materials were prepared and how are the sensors fabricated, the authors presented a useful multi-modal tactile glove with remarkable characteristics. The addition of flexion sensors compensated the tensile stretching affect on the pressure sensors. The rich tactile information was further validated in simulated tactile events.

Comments:

1. 1st paragraph: while sensory feedback from beneath the skin (joints, tendons and muscles)is closely related to skin sensory feedback, I believe that it is usually categorized as haptic.

2. The 3rd paragraph about human skin tactile sensing capability could be better placed immediately after the 1st paragraph or combined with it, leaving the remaining artificial tactile sensor background coherent.

3. Section 2.1 (Material selection for capacitive sensor dielectric) is beneficial for most researchers working in soft tactile skin sensor, especially when high sensing range and large deformation is trending. In Figure 2, smaller marker size could make it the plot representation cleaner and clearer.

4. Section 2.2: the claim that the sensitivity is two orders of magnitude higher than that of bulk elastomer pressure sensors is a bit ambitious. The authors reference [43] reported sensitivity of 1.62 MPa^-1, which is about one tenth of the sensitivity reported in this work(0.020 kPa^-1 or 20 MPa ^-1).

5. The y-axis argument "Sensitivity" in Figure 3(f) can be easily confused with that shown in Figure 3(d) and discussed in the 2nd paragraph of section 2.2. The former is the sensitivity of the relative capacitance change to tensile strain, therefore dimensionless. The authors could resolve the confusion by naming these quantities to "tensile sensitivity" and "pressure sensitivity".

6. Using solid metallic films as sensing element, these sensors undergo repeatability tests with cyclic loading described in section 2.3. It is unclear why a low pressure loading (8 kPa) is selected when the pressure sensor can measure up to 405 kPa. 100 cycles for the uniaxial tensile tests seems very different than 250,000 cycles for compression tests with little discussion.

7. Since the resistive flexion sensors has some temperature dependency, is it possible to place another reference sensor to compensate? Perhaps it could be placed on the opposite side of the current flexion sensors location, or on the side of the joints.

8. It is not clear how the compensation algorithm works in section 2.3. Specifically, how the compensating capacitive value is determined at the measured MCP and PIP joints position in equation (1) is unknown. The process of calibrating bending sensor readout to joint angle is ignored in main text, although shown in Figure S12 with poor fidelity. The compensating algorithm could be generally described as linear/quadratic and detail information presented in supplemental information.

9. In Figure 5(c) and corresponding text, the authors attempted to distinguish object stiffness using the pressure sensors. Although the results are interesting with soft polyurethane foam and rigid plastic, the difference of sensed pressure could be the result of different mass instead of stiffness. Density of polyurethane foam vary from 50 to 1000 kg/m^3, while rigid plastic is approximately 1000-1500 kg/m^3. Since the object diameters are controlled to be equal, I would suggest adding a fourth test with hollow rigid plastic cylinder to eliminate the effect of different mass.

10. The oscillation of pressure sensor readings in Figure 5(c) is very interesting and could warrant further investigation as to human grasp control capabilities and artificial control algorithm insights.

11. Figure 5(d) shows an preliminary test of the tactile glove's application potential. The task was relevant and the results are promising. Although the sample size is very much limited, I believe that a few quantitative measures can be drawn from these tests to better present the tactile glove's capability: the time the operator takes to stabilize the grasp to prescribed level, RMS of the sensor reading during stabilized grasp, initial overshot of the sensor readout. These quantitative measures could be of additional value when compared to a controlled group, where the visual tactile information is not available to the operator.

12. In section 4, while the text description was informative, graphical representation of the sensor layout is shown in the previous page on Figure 4, which might cause confusion.

13. It is unfortunate that the loading condition is not clearly described in section 4. The shape and area of the object that is applying load to the pressure sensor is unknown.