Autoposer. Drag a hand.
The body follows.

The model poses a 30-joint skeleton. Your character is almost certainly on something else, so the plugin works out which of its bones is the hips, each knee and each hand from the skeleton's shape, not from its bone names. Mixamo, Unreal and other humanoid rigs get handles without anything being renamed or rebuilt.

A rig standing in an A-pose rather than a T-pose, one exported Y-up or in centimetres, one with extra spine or twist bones: each is read as the body it is. If a bone is picked wrong, the Skeleton section of the Pose panel shows every joint and lets you choose another.

A handle that is on says “put this joint here”. A handle that is off contributes nothing, and the model places that joint from what it knows about bodies. Each handle also carries its slack: tight, and the joint goes exactly where you put it; loose, and the model may overrule it to keep the pose natural.

Three handles is the fewest it was trained to answer well. Hips, hands, feet, head and a look-at are the default set, and you can add more. The handles follow the playhead, so you can scrub to any frame and pose there.

It makes one pose at a time. There is no motion model inside it: a sequence of poses is consistent because the handles are, and the frames between your poses come from generation, not from the Autoposer.

It poses the body. Fingers get one of three hand shapes rather than being posed, and the face is left alone. On a control rig, such as Rigify or the Mixamo Control Rig, the deform bones follow the controls, so the Autoposer says so rather than starting; pose the deform skeleton instead. A body far outside ordinary proportions is posed less precisely. And it needs a humanoid: two arms, two legs and a head.