[guided]We must check that the auxiliary network can send all supply out of $s^*$ into $t^*$. By the max-flow min-cut principle, the obstruction is a cut of capacity less than the required value $B$. So we compute an arbitrary cut.
Let $Y\subseteq V$ be the set of original vertices placed on the sink side. Then the source side and sink side are
\begin{align*}
S_Y:=\{s^*\}\cup (V\setminus Y).
\end{align*}
The sink side is
\begin{align*}
T_Y:=Y\cup\{t^*\}.
\end{align*}
The arcs crossing from $S_Y$ to $T_Y$ have three types.
First, an auxiliary arc $(s^*,v)$ crosses exactly when $v\in Y$ and $\hat b(v)<0$, and its capacity is $-\hat b(v)$. These contribute
\begin{align*}
\sum_{\substack{v\in Y,\, \hat b(v)<0}}-\hat b(v).
\end{align*}
Second, an auxiliary arc $(v,t^*)$ crosses exactly when $v\in V\setminus Y$ and $\hat b(v)>0$, and its capacity is $\hat b(v)$. These contribute
\begin{align*}
\sum_{\substack{v\in V\setminus Y,\, \hat b(v)>0}}\hat b(v).
\end{align*}
Third, an original arc crosses from $V\setminus Y$ into $Y$ exactly when it belongs to $\delta^-(Y)$, and its residual capacity is $u(a)-l(a)$. These contribute
\begin{align*}
\sum_{a\in\delta^-(Y)}(u(a)-l(a)).
\end{align*}
Therefore the cut capacity is
\begin{align*}
C(Y)
=
\sum_{\substack{v\in Y,\, \hat b(v)<0}}-\hat b(v)
+
\sum_{\substack{v\in V\setminus Y,\, \hat b(v)>0}}\hat b(v)
+
\sum_{a\in\delta^-(Y)}(u(a)-l(a)).
\end{align*}
Now compare this to the required value
\begin{align*}
B=\sum_{\{v\in V:\hat b(v)>0\}}\hat b(v)
=
\sum_{\substack{v\in Y,\, \hat b(v)>0}}\hat b(v)
+
\sum_{\substack{v\in V\setminus Y,\, \hat b(v)>0}}\hat b(v).
\end{align*}
Subtracting gives
\begin{align*}
C(Y)-B
=
\sum_{a\in\delta^-(Y)}(u(a)-l(a))
+
\sum_{\substack{v\in Y,\, \hat b(v)<0}}-\hat b(v)
-
\sum_{\substack{v\in Y,\, \hat b(v)>0}}\hat b(v).
\end{align*}
The last two terms combine to $-\sum_{v\in Y}\hat b(v)$, so
\begin{align*}
C(Y)-B
=
\sum_{a\in\delta^-(Y)}(u(a)-l(a))
-
\sum_{v\in Y}\hat b(v).
\end{align*}
Thus the cut has capacity at least $B$ precisely when
\begin{align*}
\sum_{a\in\delta^-(Y)}(u(a)-l(a))\geq \sum_{v\in Y}\hat b(v).
\end{align*}
It remains to show that this is the same inequality as the one in the theorem. Using the definition of $\hat b$,
\begin{align*}
\sum_{v\in Y}\hat b(v)
=
\sum_{v\in Y}b(v)
-\sum_{v\in Y}\sum_{a\in\delta^-(\{v\})}l(a)
+\sum_{v\in Y}\sum_{a\in\delta^+(\{v\})}l(a).
\end{align*}
An arc with both endpoints in $Y$ appears once in the incoming lower-bound sum and once in the outgoing lower-bound sum, so it cancels. An arc entering $Y$ from $V\setminus Y$ appears only in the incoming lower-bound sum and therefore contributes $-l(a)$. An arc leaving $Y$ appears only in the outgoing lower-bound sum and therefore contributes $+l(a)$. Hence
\begin{align*}
\sum_{v\in Y}\hat b(v)
=
\sum_{v\in Y}b(v)
-\sum_{a\in\delta^-(Y)}l(a)
+\sum_{a\in\delta^+(Y)}l(a).
\end{align*}
Substituting this into the residual cut inequality gives
\begin{align*}
\sum_{a\in\delta^-(Y)}(u(a)-l(a))
\geq
\sum_{v\in Y}b(v)
-\sum_{a\in\delta^-(Y)}l(a)
+\sum_{a\in\delta^+(Y)}l(a),
\end{align*}
and adding $\sum_{a\in\delta^-(Y)}l(a)$ to both sides yields
\begin{align*}
\sum_{a\in\delta^-(Y)}u(a)-\sum_{a\in\delta^+(Y)}l(a)
\geq
\sum_{v\in Y}b(v).
\end{align*}
This is exactly the assumed cut condition.[/guided]