Picture a book resting on a table. In the free body diagram the table, the floor and you vanish; only the book remains, drawn as a box or a dot. Then ask what touches it and what acts at a distance. Earth pulls it down: its weight, mg. The table touches it and pushes up: the normal force. Nothing else touches it, so the diagram has exactly two arrows, equal in length because the book is at rest.
Each arrow answers the question "what outside thing pushes or pulls this object?" The book also presses down on the table, but that force acts on the table and belongs on the table's diagram. A surface in contact gives a normal force and possibly friction; a rope gives tension along the rope; gravity is the one everyday force that needs no contact. If you cannot name an arrow's source, the arrow should not be there: "the force of motion" and "centrifugal force" fail this test.
Contrast a free body diagram with a sketch of the whole scene. A drawing of a person dragging a sled up a hill, with arrows on the person, the rope and the sled, is a picture of several objects; adding its arrows mixes forces that act on different bodies. The net force is never drawn either: it is the sum you compute from the arrows. For a box sliding down a rough ramp, the correct diagram has three arrows: weight straight down, normal force perpendicular to the ramp, friction up the slope.
Think of it like this. A free body diagram is a passport photo of one object: cropped from the crowd, with nothing else in frame. Every arrow is a hand reaching in from outside the frame, and you show only the hands reaching in, never the ones the object reaches out with. Gravity is the one hand that reaches in without touching.
Core idea. Isolate one object, draw every force that something else exerts on it and none that it exerts on others, then add those arrows as vectors; that sum alone decides the object's acceleration.