Intrinsic Possibilities, Extrinsic Possibilities, and Possible Worlds Analyses of Modality

On a possible-worlds analysis of possibility, for some x to be possible is for x to exist at at least one possible world. From this, it follows that x is impossible just in case there is no possible world at which x exists.

Here's my worry: such an account is too coarse-grained to distinguish between intrinsic and extrinsic impossibilities; in particular, it's too coarse-grained to handle putative cases of entities that are intrinsically possible, yet extrinsically impossible. Thus, consider the following thought experiment:

Goody and Baddy
Consider two beings, Goody and Baddy. Goody is an Anselmian being, and Baddy is an inherently malevolent being who gets his kicks by torturing all other creatures that happen to exist. However, in virtue of his essential goodness, Goody is inherently such that he prevents the existence of Baddy in any world in which Goody exists. Thus, since Goody is an inherent Baddy-blocker, and since Goody exists in all possible worlds, there is no world at which Baddy exists.

If Goody blocks Baddy's existence at every world, then Baddy's impossibility is due to factors extrinsic to it. By contrast, consider a round square, RS. Like Baddy, RS exists at no possible world. However, unlike Baddy, RS fails to exist at any world in virtue of its own nature.

So a worlds analysis of possibility appears to be too coarse-grained to handle cases of entities that are intrinsically possible, yet extrinsically impossible; therefore, I worry that a straight possible-worlds analysis fails to capture the essence of possibility.