Would you please once more describe clearly what you are trying to do. And this time don't say that you want to design a color sensor. Instead say what problem you want to solve. What would the sensor readings be used for? Why? What would happen if the sensor reading is off in luminanse (intensity) or chrominanse (color)? Would the sensor have to be able to detect IR or UW? Must it be able to work both in direct sunlight, and under the stars a night without a moon?

A color sensor is a sensor that decides how much red/green/blue there is, or that detects the amount of energy in very specific wavelength bands.

That means that a color sensor must have a calibrated sensitivity in calibrated wavelength regions.

It doesn't matter if you get a flashlight lamp, a LED etc. You will still not manage to calibrate any sensor. All you will know, is how much the sensor will respond to the light from the lamp. But you will not know how much the sensor will respond to light not emitted by the lamp. Most times, you will not even know the spectral distribution of the light from the lamp.

There are many ways to detect if it is light or dark in a room or outside with a reasonable precision. But measuring color on the other hand is a very hard problem to solve. Where would you get color filter from? They may be massproduced for a low cost, but buying them in small quantity means paying a lot of money.

But getting a good color filter isn't enough. The color filter must be matched with a sensor with a reasonable sensitivity within the spectral range that the filter sends through. And when you have both a filter and a sensor, then you need calibrated light of all visible wavelengths to allow you to compare the amount of light you shine on your sensor, with the sensor reading, allowing you to create a calibration table for the sensor. If the sensor design can be built in multiple copies with a small deviation between the individuals, then you may use a single calibration table. If the sensor has a lot of deviation, then each produced sensor will need its own calibration table, which means that whoever builds such a sensor must have access to calibrated light for the full visible spectrum.

In the end, a color sensor may be one of the harder things you can ever try to build, even if a final product may be tiny and look trivial. That is the reason why you would do better actually buying a sensor from a company who have spent significant amounts of time and money on the design. And that is also a reason why a redneck low-end solution may be to try a web-cam or similar, where you have a CMOS or CCD sensor with three color filters reasonably matched to the sensitivity of our eyes. It isn't good, but it would take a significant amount of work to just get close to the capability of a cheap web-cam.

That is a very good reason why you should care, if someone do post a link to a cheap sensor intended for measuring color.