can i perform skin detection using a set of rules (things like x1 < Y < x2 )
The short answer is yes, you can.
However, luminance (Y) is actually irrelevant. It's the chrominance (CbCr or UV) that matters.
One of the most cited papers in this area is this one (uncompressed PS file). I've implemented it and it seems to work fairly well.
UPDATE: the link above seems to have become broken, so here is the bibtex citation of the paper:
@ARTICLE{767122,
author={Chai, D. and Ngan, K.N.},
journal={Circuits and Systems for Video Technology, IEEE Transactions on}, title={Face segmentation using skin-color map in videophone applications },
year={1999},
month={jun},
volume={9},
number={4},
pages={551 -564},
keywords={H.261-compliant coder;chrominance component;complex background scene;face-segmentation algorithm;fast algorithm;foreground/background coding;head-and-shoulders view;human skin color;input image;luminance;perceptual quality;pixels;regularization processes;reliable algorithm;simulation results;spatial distribution characteristics;test images;universal skin-color map;video coding;videophone applications;videophone sequence;videotelephony;brightness;image colour analysis;image segmentation;image sequences;video coding;videotelephony;},
doi={10.1109/76.767122},
ISSN={1051-8215},}