It's good (apart notational slips) until you start putting in $\delta$.
You have to show that the inequality $|x|^2+|x|<\varepsilon^2$ is satisfied for $|x|<\delta$ for some $\delta>0$ to determine.
You needn't find the better $\delta$, just one that does the job. So you can restrict yourself to $|x|<1$, which implies $|x|^2<|x|$. So, if you satisfy $|x|+|x|<\varepsilon^2$, you're done; but this is
$$
|x|<\frac{\varepsilon^2}{2}
$$
so you can take
$$
\delta=\min\left\{\frac{\varepsilon^2}{2},\frac{1}{2}\right\}
$$
(because, remember, you want to ensure $|x|<1$).
A more complicated approach would be to solve the inequality
$$
|x|^2+|x|-\varepsilon^2<0
$$
that is satisfied for
$$
\frac{-1-\sqrt{1+4\varepsilon^2}}{2}<|x|<\frac{-1+\sqrt{1+4\varepsilon^2}}{2}
$$
which is the same as
$$
|x|<\frac{-1+\sqrt{1+4\varepsilon^2}}{2}
$$
because the least root is negative.
Therefore you could take
$$
\delta=\frac{-1+\sqrt{1+4\varepsilon^2}}{2}
$$