The only guidelines that I have seen are internal ones. And that is my first recommendation, you should have a method validation SOP and a validation protocol specific for the method validation that you are to perform. More on that can be found in a previous discussion:
 http://www.sepsci.com/chromforum/viewtopic.php?t=7705
Are you looking at the API assay only or are you testing drug products? The experiment will be varied a bit depending on which you are doing.
Perform the stress tests on a solution as that is where, in general, you will get more degradation reactions to occur. Obviuously, you will use solutions for the acid, base and oxidation tests, but also use solutions for the heat and light stress tests.
If you are testing a drug product, then you will want the excipeints in the solution along with the API. Also test a solution of the placebo for each stress condition. Knowing degradation of the excipients is very useful.
Use some sceintific guidance/guess work here based on what you know about your API. Do you have any idea about its stability? You may need to do trial and error experiments: try a stress condition, if you get too much degradation then repeat but with a less stressful condition.
Use 0.1 N HCl, 0.1 N NaOH, 10% H2O2, 60 degrees C and a light cabinet. Make solutions as noted above. You can choose the amount of stress solution to add and you can even add some of your method sample diluent (which is what you would use for the heat and light samples). Store the acid, base and oxidation stress samples for whatever time period that is appropriate based on your knowledge of the API; it could be 4 hours, it could be overnight. The light cabinet does not need to be calibrated or for any given lux, just put the solution samples in the cabinet for 24 hours. Choose your heat storage time: 4 hours, overnight, etc.
There was a previous discussion for how to stop the peroxide reaction:
 
http://www.sepsci.com/chromforum/viewtopic.php?t=7068
Lastly, be catious to use a range limit for the amount of degradation. What happens if you don't get 5% degration in light because your API is not light sensitive? ANSWER: Your QA makes you write a deviation report to explain why or they get you to repeat the experiment until you do get 5% degradation (which may never happen). Having an upper limit only makes sense because, in the real world, a sample with more than 10% degradation is a failure and you would never need to analyze such a sample. Write your protocol carefully, or QA will get you.
Regards,
Dan