Hi,
I'd like to stress the fact that Thomas used equations and rules, but not assignments like F[a_, b_] := ... because then trying to insert F[a, b] in the computations would immediately use that assignment.
Thomas' idea is to replace CD[a][A[b]] by a sum of its antisymmetric part (i.e. F[a, b]) and its symmetric part. Another possibility is to define a rule of the form:
rule = CD[a_][A[b_]] :> F[a, b] + CD[b][A[a]] /; IndexOrderedQ[{a, b}]
Then you also have
In[ ]:= CD[-b][CD[-a][A[b]]] - CD[-b][CD[b][A[-a]]] /. rule
Out[ ]= CD[-b][F[-a, b]]
Cheers,
Jose.