I'm quite surprised, people are asking for tools to develop their own skins for other NES games. HDNES exists for about 4 years now, but I haven't seen many graphic packs developed for it since then.
There is no problem giving away that FCEUX, here it is: https://yadi.sk/d/GRCG_Zaq3JKCaA
But this was made as a quick and dirty hack. Maybe it's preferrable to stick to standards? Now as Mesen supports HDNES packs, maybe it is better to use them for facelifting NES games?
They allow for upping tile resolution, and they even seem to replace background music (what people asked here).
My hack, on the other hand, due to its nature won't be able to do that. Speaking of its other limitations, it only supports 32000 colors (5 bits R, G, B) in replacement patterns. It also can not replace "something with nothing" (if an original sprite has any non-transparent color in a particular row, a replacement sprite should be non-transparent in that row too). Another limitation is that it most likely won't support fadein/fadeout, and certainly does not support NES color emphasizing and B/W mode.
And finally the most important one: the hack does not replace the tileset entirely. It only checks if a particular original pattern (a row of 8 pixels) is read from a particular address in a pattern table (a special area of NES VRAM). If it detects such an event, it renders a replacement pattern instead. So two (or more) equal original patterns can't be replaced with different replacement patterns. When I started my work, I thought this to be a minor issue. I was wrong. Despite Castlevania being a relatively small game, it happened all the time. I had to develop a concept of trigger patterns. If a particular (trigger) row of 8 pixels is detected early in a frame (preferrably in a top row), then a special subset of so called triggered patterns is used for replacement until the end of a frame. As triggers I mostly used regular color filled "night sky" or abundant brick patterns.
I'm afraid this concept won't work in a tile rich game like Castlevania 3. The non-linearity of levels is not a problem here. The problem is the total amount of the tiles in the game. It may turn out impossible to choose a trigger in certain levels.
Another problem with Castlevania 3 is that, as far as I remember, it uses a lot of custom effects like fog or water. How will they look like with a skin hack is impossible to imagine.
Although I admit, it is fun to imagine Trevor from SotN making his way through the original Castlevania 3.