[proofplan]
The proof is a restriction of the RSK bijection to one fixed shape. The RSK correspondence identifies each permutation $w\in S_n$ with an ordered pair $(P(w),Q(w))$ of standard Young tableaux of the same shape. The condition $\operatorname{sh}(w)=\lambda$ is exactly the condition that both tableaux have shape $\lambda$, so the desired fibre is in bijection with ordered pairs of standard Young tableaux of shape $\lambda$. Counting the first and second tableau independently gives $(f_\lambda)^2$.
[/proofplan]
[step:Restrict the RSK bijection to permutations of shape $\lambda$]
Let
\begin{align*}
A_\lambda:=\{w\in S_n:\operatorname{sh}(w)=\lambda\}
\end{align*}
be the set of permutations whose RSK shape is $\lambda$. Let
\begin{align*}
\operatorname{SYT}(\lambda):=\{T:T \text{ is a standard Young tableau of shape } \lambda\}
\end{align*}
be the set of standard Young tableaux of shape $\lambda$, so that $f_\lambda=|\operatorname{SYT}(\lambda)|$ by definition.
Let
\begin{align*}
\mathcal{C}_n:=\{(P,Q):\text{there exists }\mu\vdash n\text{ such that }P,Q\in\operatorname{SYT}(\mu)\}
\end{align*}
denote the set of ordered pairs of standard Young tableaux of the same shape and total size $n$. By [citetheorem:8435], the map $\operatorname{RSK}:S_n\to \mathcal{C}_n$ given by $w\mapsto (P(w),Q(w))$ is a bijection. Since $\operatorname{sh}(w)$ is the common shape of $P(w)$ and $Q(w)$, the restriction of $\operatorname{RSK}$ to $A_\lambda$ has image precisely
\begin{align*}
\operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda).
\end{align*}
Therefore the restricted map
\begin{align*}
\operatorname{RSK}|_{A_\lambda}:A_\lambda\to \operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda)
\end{align*}
is a bijection.
[guided]
We isolate the fibre whose cardinality we want to compute. Define
\begin{align*}
A_\lambda:=\{w\in S_n:\operatorname{sh}(w)=\lambda\}.
\end{align*}
Thus the left-hand side of the desired identity is $|A_\lambda|$.
Next define
\begin{align*}
\operatorname{SYT}(\lambda):=\{T:T \text{ is a standard Young tableau of shape } \lambda\}.
\end{align*}
By the meaning of $f_\lambda$, we have
\begin{align*}
f_\lambda=|\operatorname{SYT}(\lambda)|.
\end{align*}
Now define
\begin{align*}
\mathcal{C}_n:=\{(P,Q):\text{there exists }\mu\vdash n\text{ such that }P,Q\in\operatorname{SYT}(\mu)\}
\end{align*}
for the set of ordered pairs of standard Young tableaux of the same shape and total size $n$. Apply the RSK bijection for permutations, [citetheorem:8435]. Its hypotheses are satisfied because we are working with permutations in $S_n$, and the theorem states that the map $\operatorname{RSK}:S_n\to\mathcal{C}_n$ given by $w\mapsto (P(w),Q(w))$ is a bijection.
The only point to check is what the condition $\operatorname{sh}(w)=\lambda$ becomes after applying RSK. By definition, $\operatorname{sh}(w)$ is the common shape of the insertion tableau $P(w)$ and the recording tableau $Q(w)$. Hence
\begin{align*}
\operatorname{sh}(w)=\lambda
\end{align*}
holds exactly when both $P(w)$ and $Q(w)$ have shape $\lambda$. Therefore the RSK image of $A_\lambda$ is precisely
\begin{align*}
\operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda).
\end{align*}
Because the original RSK map on $S_n$ is bijective, its restriction to a subset is injective, and every pair in the displayed product has a unique preimage in $S_n$. Since that pair has shape $\lambda$, its preimage lies in $A_\lambda$. Thus
\begin{align*}
\operatorname{RSK}|_{A_\lambda}:A_\lambda\to \operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda)
\end{align*}
is a bijection.
[/guided]
[/step]
[step:Count ordered pairs of standard tableaux of shape $\lambda$]
Since $\operatorname{RSK}|_{A_\lambda}$ is a bijection, finite sets have equal cardinality under it:
\begin{align*}
|A_\lambda|=|\operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda)|.
\end{align*}
The two coordinates of an ordered pair are chosen independently from the same finite set $\operatorname{SYT}(\lambda)$, so
\begin{align*}
|\operatorname{SYT}(\lambda)\times \operatorname{SYT}(\lambda)|=|\operatorname{SYT}(\lambda)|^2.
\end{align*}
Using $|\operatorname{SYT}(\lambda)|=f_\lambda$, we obtain
\begin{align*}
|\{w\in S_n:\operatorname{sh}(w)=\lambda\}|=|A_\lambda|=(f_\lambda)^2.
\end{align*}
This is the required identity.
[/step]