As far as I understand, the adapter pattern is creating a wrapper object for our real object of interest, simply one more level of indirection, which provides flexibility. The flexibility is in that if the real object's interface is changed, then we change the wrapper interface pointing at the real object, leaving the client-side exposed interface unchanged.
The proxy pattern is the same, with the difference that every proxy wrapper provides only a coherent subset of the real object's functionality. Why would this be useful when we strive to make "one class for one purpose" is beyond me.
Have I gotten this correctly?