It is relatively easy for Bacillus thuringiensis subspecies. By heating the soil samples all bacteria which don’t forming spores will be killed. After streaking the bacteria, washed from the soil samples, on agar the growing B. thuringiensis colonies can be selected by their specific shape and finally by the identification of crystal inclusion bodies in the cells of several day old cultures. You will find more detailed descriptions and more sophisticated methods in the internet. However, for the identification of the subspecies you will need to carry out PCR protocols with specific primers for the different delta-endotoxins (check internet for ‘characterization thuringiensis soil’ and focus on publications not older than 1998).
No that will not work properly. The density of the entompathogenic bacteria in the soil will be usually too low for a bioassay. You need to propagete the isolated bacteria before you can use them in a bioassay.
On the other hand you can expose the catarpillars (e. g. Galleria) directly to the soil. In this way you should be able to catch some entomopathogenic fungi. But this methods doesn't work well for bacteria.
You need to be very specific. For example instead of searching all kind of microbes, you need to concentrate on fungi or bacteria. Also what insect you are targeting?
I feel you can try baiting method using your target insect with moist soil sample.
As Ahmed Al-Rawahi told you, need to be more specific. I suggest you check out for entomopathogenic fungi and try to isolate them from dead insects itself in the proper media culture. From the soil, that is correct you could find effective potential new isolates. B. bassiana is quite good as M. anisopilae
Try to get the pure isolates through the selective media. Do the pathogenicity test from your several isolated strains from the different origins. The potentiall strain normally will provide mortality as effective as the chemical insecticide (i.e cypermethrin, trichlorphon etc) but a bit lower and slower. Based on my experiences >80% corrected mortality in a glass house study should be a very potential strain for the next future of mass production program.