The statement holds true only for distinct roots $\rm\, a\ne b.\:$ As such, the proof will require use of $\rm\, a\ne b,\:$ e.g. by cancelling $\rm\,a-b.\:$ Let $\rm\,f(x)\,$ be the quartic and let $\rm\,g(x)\,$ be the sectic. Then $\rm\,g(ab)\,$ can be reduced to $\,0\,$ by using $\rm\,\color{#C00}{f(a)} = 0 = \color{#C00}{f(b)}\ $ and $\rm\ \color{#C00}{h(a,b)} = \dfrac{f(a)-f(b)}{a-b} =\, 0,\,$ namely
$$\rm\begin{eqnarray} g(ab)\,\ &=&\ \rm (a^5\!+ a^4b+ a^3b^2\!+\!a^2 b^2\!+a^3\!+1)\ \color{#C00}{f(a)} \\
&& \ + \rm\ (a^6 b^2\!-\!a^6 b+a^6\!+a^4)\ \color{#C00}{f(b)}\\
&&\ -\ \rm (a^6\!+a^4\!-\!a^3)\ \color{#C00}{ h(a,b)}\\
&=&\ \ 0\quad\rm
\end{eqnarray}$$
Remark $\ $ There may well be a simpler such expression (e.g. a symmetric one). I have not attempted to simplify it.