In any category π, an object T is a terminal object if and only if it has the property: for each object X in π there is exactly one map from X to T.
See the earlier definition of terminal object on p. 213.
Theorem (p. 229)
Suppose that π is any category and that both T_1 and T_2 are terminal objects in π. Then T_1 and T_2 are isomorphic; i.e. there are maps {f : T_1 \rightarrow T_2}, {g : T_2 \rightarrow T_1} such that {g \circ f} is the identity of T_1 and {f \circ g} is the identity of T_2.
Proof (p. 229)
To show that T_1 and T_2 are isomorphic we need first of all a map {T_1 \rightarrow T_2}. Because T_2 is terminal there is a map {f : T_1 \rightarrow T_2}. We need a map {g : T_2 \rightarrow T_1}. Again there is one because T_1 is terminal. But this does not prove yet that T_1 is isomorphic to T_2. These two maps have to be proved to be inverse to each other. Because T_1 is terminal, there is only one map from T_1 to T_1. Therefore, {g \circ f = 1_{T_1}}. [Because T_2 is terminal, there is only one map from T_2 to T_2. Therefore, {f \circ g = 1_{T_2}}.]
We provided a proof of this theorem in Article IV using the mathlib API for terminal objects. We repeat that proof here, but this time using only the book definition.