In case 'indoor sensing' refers to gas sensing: Just a few molecules absorb strongly in the visible such as NO2. Using e.g. multi-passing to create sufficient path length it might be possible to detect such molecules in sufficiently polluted indoor environments.
Other solutions are possible as for example discussed in http://www.merl.com/publications/docs/TR2005-097.pdf
The problem is that in lighting applications the LED is blue emitter illuminating a white phosphor. To use the same structure as an optical receiver means you have a large loss in front of your detector.
In the case of individual RGB emitters without phosphors, you still have the problem of operating the diode as a detector, where you would want the diode reverse biased, not forward biased as an emitter. In principle, you could alternate the bias.
New type of LEDs for functioning both emitter and receiver is an interesting topic. By changing between reverse and forward biased, one can use two LEDs for full-duplex communication.