Characters get the attention; the majority of moving objects in a game are doors, lids, levers, wheels and machinery. They need much less.
Hierarchy is usually enough
A door needs a pivot at its hinge and a rotation. A wheel needs a pivot at its axle. Most prop animation is a transform on a correctly-placed origin — no skeleton, no skinning, no weights.
Getting the pivot right at modelling time is the whole job. A door whose origin is at its centre cannot be opened by rotating it.
Parenting for compound objects
A crane is a base, a tower, a jib and a hook, each parented to the last. Rotating the tower carries everything above it. This is the same hierarchy idea as a skeleton, without the mesh deformation.
For this to work, the model has to be built in parts with separate origins — which is where segmentation helps, and where its limits show: segmented parts have no hidden geometry behind their cut, so a drawer that slides out reveals a hole.
When a prop needs bones
- Anything that bends. Cables, hoses, cloth, chains.
- Anything with many pieces where a bone hierarchy is more manageable than an object hierarchy.
- Anything sharing an animation system with characters, for pipeline consistency.
Physics instead of animation
Some prop motion is better simulated than animated — debris, cloth, ropes, swinging signs. Simulation reacts to the world; animation does not. The trade is control and cost.
Practical advice
Decide at modelling time. Where the pivots go, which parts are separate, whether the interior is ever revealed — all of these are cheap decisions before the model is finished and expensive ones after.