It's entirely possible. Preferably you just attach them to the game object representing that part of the body. If your mesh didn't import into different game objects then you should look into how to do that with import settings, as the alternative is to make gameobjects that are invisible with the colliders you want, then animate them with your model to match. But that is VERY tedious.
It's also important to know that a parent rigidbody acts like a vacuum for collision messages. Any colliders will belong to the nearest parent rigidbody, and work together as such. Make sure you have a rigidbody on your character, and the colliders are on child gameobjects of that rigidbody. They will also however collide with eachother, so you should put them on a collision layer and use the Collision Matrix to make them avoid self collision.
Lastly, the reason we use a capsule collider is because it avoids a lot of edge cases to simplify our collision geometry. A common thing to do is seperate our physical body collier from our damage/hitbox. Our physical collider should be simple, round, and aproximate. Thus a capsule collider is perfect. However for say, first person shooters, we want precise boundaries for bullet detection. Here we can go and add a seperate set of colliders, and again put them on a layer that only colliders with things that damage it. These colliders are more precise but don't handle physics, merely collision / hit detection. You should probably even make them triggers.