I'm having trouble proving a composite of $\mathcal{C}^k$ functions is $\mathcal{C}^k$ without use of partial differentiation. To set up my question, consider the following $\mathcal{C}^k$ functions defined on open sets:
$\begin{matrix} U & \xrightarrow{f} & V & \xrightarrow{g} & \mathbb{R}^p, \\ \cap & & \cap & & \\ \mathbb{R}^n & & \mathbb{R}^m\end{matrix}$
To prove the composite $g \circ f:U \to \mathbb{R}^p$ is $\mathcal{C}^k$, it is usually suggested to start by considering a single application of chain rule:
$D(g \circ f)(p) = Dg(f(p)) \circ Df(p). \tag{1}$
I would like to then apply the chain rule again, but the operation $\circ$ above both is and isn't function composition. To explain, it is in the sense that if
If $\,\,\, \alpha = Df(p) \in L(\mathbb{R}^n,\mathbb{R}^m), \\ \,\, \beta = Dg(f(p)) \in L(\mathbb{R}^m,\mathbb{R}^p), \\ \,\, \gamma = D(g \circ f)(p) \in L(\mathbb{R}^n,\mathbb{R}^p) \\ \mathrm{then\,\,} \gamma = \beta \circ \alpha,$
but also isn't, because upon writing
$A = Df: U \to L(\mathbb{R}^n,\mathbb{R}^m) \\ B = (Dg) \circ f : U \to L(\mathbb{R}^m,\mathbb{R}^p) \\ C = D(g \circ f): U \to L(\mathbb{R}^n,\mathbb{R}^p),$
$(1)$ becomes $C(p) = B(p) \circ A(p)$, and $C = B \cdot A$, where $\cdot$ is a form of multiplication rather than function composition ($B \circ A$ makes no sense, as their source and target spaces make them incompatible with function composition). Apparently, a form of product rule is needed to find the total derivative of $C = B \cdot A$.
I'm aware that both $A,B$ are Frechet differentiable (presuming $k>1$), but I'm not aware of a product rule regarding total differentiation of anything other than a product of real valued functions. Is there such a rule?
I suspect there is a higher order chain rule as well, yet that which uses a product rule for certain vector valued functions.