One approach to solve this without any advanced theorems is to start with a guess like:
$$P(x) = A_0 + A_1x^1 + A_2x^2 + A_3x^3 + A_4x^4$$
Then expand
$$P(\sqrt{2} + \sqrt{3} + \sqrt{6}) = 0$$
and you'll some equation of the form:
$$B_0 + B_1 \sqrt{2} + B_2 \sqrt{3} + B_3 \sqrt{6} = 0$$
where none of the $B$ contain any radicals, they are just linear combinations of the $A_k$. In this case it is that:
$$\begin{align}
B_0 &= A_0 + 11~A_2 + 36~A_3 + 265~A_4 \\
B_1 &= A_1 + 6~A_2 + 29~A_3 + 180~A_4 \\
B_2 &= A_1 + 4~A_2 + 27~A_3 + 136~A_4 \\
B_3 &= A_1 + 2~A_2 + 21~A_3 + 92~A_4 \\
\end{align}$$
If you take it for granted, that there are no integers $m, n$ such that $m\sqrt{x} + n\sqrt{y} = 0$ except for $m = 0$ and $n = 0$ (for sufficiently different $x$, $y$), then since $P(x) = 0$ then
$$\begin{align}
0 &= A_0 + 11~A_2 + 36~A_3 + 265~A_4 \\
0 &= A_1 + 6~A_2 + 29~A_3 + 180~A_4 \\
0 &= A_1 + 4~A_2 + 27~A_3 + 136~A_4 \\
0 &= A_1 + 2~A_2 + 21~A_3 + 92~A_4 \\
\end{align}$$
That's 4 equations and 5 variables. Naturally you'll have an extra variable, since if $P(x)$ is a solution, then so is $n~P(x)$. Since the question only asked if such a polynomial exists, you could have skipped most of the calculations and just pointed out that such calculations were possible without actually doing them. If you do bother to do the calculations, the result is $P(x) = n(23 + 48~x + 22x^2 + x^4)$ for any integer $n$.
Also, the reason there are only 4 values in the $B$ array is because $\sqrt{2}\sqrt{3} = \sqrt{6}$, so that limits the number of radical coefficients after the expansion of $P(\sqrt{2} + \sqrt{3} + \sqrt{6})$. If we had expanded something like $P(\sqrt{2} + \sqrt{3} + \sqrt{11})$, then there would have been $2^3$ instead of $2^2$ terms in the expansion.