I had a piece of code which did all vertex-transformations (bones/world, view, projection) of my animated models on the cpu CPU. During these transformations I automatically got the geometry to build the shadow volume from. To use the advantages of hardware T&L I have now converted the animated model to a skinned mesh(D3DXMESHCONTAINER + D3DXFRAME hierachy) to get the transformations to the gpu (using indexed vertex blending for rendering). the problem now is that I don't have access to the transformed vertices any longer which I need to construct the shadow volume :-/ What would be the best way to solve this problem? Transforming the vertices a second time on the cpu would reduce the whole process to absurdity I guess. All examples of shadow volumes I've found so far are either using static objects or transform the whole model on the cpu (the way I've done it before) Isn't there a better solution?

You could use vertex blending to do shadow volume extrusion. Flipcode has an article about this. I'm not sure how well this would work with skinned meshes, though.

Alternatively, and this is what I suggest, you could use vertex shaders to do both.