One thing to do is relax the problem. Instead of the problem posed, try solving the following problem:
\begin{align}
\text{minimize}\hspace{10pt} & 2x+2y-z \\
\text{subject to}\hspace{10pt} & x + y + z = 10.
\end{align}
This can be solved using Lagrange multipliers. Briefly, the point that minimizes $f(x,y,z)=0$ subject to $g(x,y,z)=0$ is the point where $\nabla f(x,y,z) = \lambda \nabla g(x,y,z)$. That is, the point that optimizes the problem is the one where the gradient of $f$ is a scalar multiple of the gradient of $g$.
Once you solve that problem, you'll see that the value that minimizes it also satisfies $x\geq 0$, $y\geq 0$, and $z\geq 0$. If a point solves a relaxed problem then it also minimizes the original problem. But, since it satisfies the additional constraint, it must minimize the original problem. So you're done!