[guided]The only possible obstruction to defining $A([p,f])=[p,a(f)]$ is that the same point of $P\times_G F$ may have many representatives. We therefore compare two representatives that are related by the defining equivalence relation.
Let $q:P\times F\to P\times_G F$ and $q':P\times F'\to P\times_G F'$ denote the quotient maps. Define $\widetilde A:P\times F\to P\times_G F'$ by
\begin{align*}
\widetilde A(p,f):=[p,a(f)].
\end{align*}
To descend $\widetilde A$ to the quotient, we must show that $\widetilde A(p,f)$ has the same value after replacing $(p,f)$ by an equivalent pair.
Assume $(p_2,f_2)=(p_1g,g^{-1}\cdot f_1)$ for some $p_1\in P$, $f_1\in F$, and $g\in G$. The equivariance hypothesis on $a:F\to F'$ says precisely that $a(h\cdot f)=h\cdot a(f)$ for every $h\in G$ and $f\in F$. Applying this with $h=g^{-1}$ gives
\begin{align*}
a(g^{-1}\cdot f_1)=g^{-1}\cdot a(f_1).
\end{align*}
Therefore
\begin{align*}
\widetilde A(p_2,f_2)=[p_1g,a(g^{-1}\cdot f_1)].
\end{align*}
Substituting the equivariance identity gives
\begin{align*}
[p_1g,a(g^{-1}\cdot f_1)]=[p_1g,g^{-1}\cdot a(f_1)].
\end{align*}
Now the right-hand side is exactly one of the pairs identified with $(p_1,a(f_1))$ in the associated bundle $P\times_G F'$, because the equivalence relation there identifies $(p',f')$ with $(p'g,h^{-1}\cdot f')$ for group elements acting on the fibre. Thus
\begin{align*}
[p_1g,g^{-1}\cdot a(f_1)]=[p_1,a(f_1)].
\end{align*}
Hence $\widetilde A(p_2,f_2)=\widetilde A(p_1,f_1)$. This proves that $\widetilde A$ is constant on equivalence classes, so it descends uniquely to a map $A:P\times_G F\to P\times_G F'$ with $A([p,f])=[p,a(f)]$.[/guided]