You cannot, because your assertion is false. Consider $n=3,\ p=2$ and $M=\operatorname{diag}(1,1,0)$. For any $i\ge1$, we have $M^i=M$ and $\operatorname{tr}(M^i)=\operatorname{tr}(M)=0$. Yet $M+I=\operatorname{diag}(0,0,1)$ is not invertible.
Your assertion is true, however, if you replace the condition $p \nmid n$ by $p=0$ or $p>n$. To prove the assertion, you may modify JBC's argument to a related question. The key is that $M$ must be nilpotent. Suppose the contrary. Then $M$ has, over the algebraic closure of $K$, a set of distinct non-zero eigenvalues $\lambda_1,\ldots,\lambda_r$. Let $n_i$ the multiplicity of $\lambda_i$. From the condition that $\operatorname{tr}(M^i)=0$ for $i=1,2,\ldots,n$, we get
$$
\underbrace{\pmatrix{
\lambda_1&\lambda_2&\cdots&\lambda_r\\
\lambda_1^2 & \lambda_2^2 & \cdots & \lambda_r^2\\
\vdots & \vdots & \vdots & \vdots\\
\lambda_1^r & \lambda_2^r & \cdots & \lambda_r^r
}}_V
\pmatrix{n_1 \\ n_2 \\ \vdots \\ n_r}
=\pmatrix{0 \\ 0\\ \vdots \\ 0}.\tag{1}
$$
The square matrix $V$ on the left is equal to
$$
\pmatrix{
1 & 1 & \cdots & 1 \\
\lambda_1&\lambda_2&\cdots&\lambda_r\\
\lambda_1^2 & \lambda_2^2 & \cdots & \lambda_r^2 \\
\vdots & \vdots & \vdots & \vdots \\
\lambda_1^{r-1} & \lambda_2^{r-1} & \cdots & \lambda_r^{r-1}
}
\pmatrix{\lambda_1\\ &\lambda_2\\ &&\ddots\\ &&&\lambda_r},
$$
which is the product of a Vandermonde matrix and a diagonal matrix. As all $\lambda_i$s are nonzero and distinct, $V$ is invertible. Yet, as $p>n$, the vector $(n_1,\ldots,n_r)^T$ is nonzero. So, we arrive at a contradiction in $(1)$. Hence $M$ must be nilpotent and in turn $M+I$ is invertible.