Use a blue photo filter in the front of a light sensor. Placed in a parabolic mirror.
Use two filters one more blue (:-)) than the other in front of two light sensors.
You will not be able to find the spectral signature, but at least you can say something about the blue color of the sky, using differences in the intensity measured by two sensors. This way there is no need to convert arbitrary values into lux or similar.
To reproduce the color, use two white leds commanded by the two signals logged by the two sensors. It is not going to be accurate, but is fun and conclusive (to a certan extent).