I would tend to think about this in terms of the following fact.
Fact: The minimum polynomial $p$ of a linear transformation $T: V \to V$ always satisfies $\deg(p) \leq \dim(V)$, and equality holds precisely when $T$ is cyclic.
If $T:V \to V$ is a linear transformation and $W \subset V$ is a $T$-invariant subspace, then $T$ descends to a map on the quotient $V/W \to V/W$. Let $p_V,p_W$ and $p_{V/W}$ denote the minimum polynomials of $T$ itself, the restriction of $T$ to $W$, and the map $T$ induces on $V/W$. Note that
$$\begin{align}\deg(p_W) + \deg(p_{V/W}) \geq \deg(p_V) && && (1)\end{align}$$
because $p_W \cdot p_{V/W}$ is a monic polynomial of degree $\deg(p_W) + \deg(p_{V/W})$ annihilating $T$ whereas $p_V$ is, by definition, the unique monic polynomial of smallest possible degree annihilating $T$. To see $p_W \cdot p_{V/W}$ annihilates $T$, recall $p_{V/W}(T)$ is zero, as a map of $V/W$, and so $p_{V/W}(T) v \in W$ for any $v \in V$. Thus, $p_W(T) \cdot p_{V/W}(T) v = 0$ for any $v \in V$, and we are done.
If $T$ is cyclic, then the map it induces on $V/W$ is clearly cyclic too; if $v \in V$ is cyclic for $T$, then $v+W$ is cyclic for the map on the quotient. So, from the fact and the inequality (1), we get
$$\deg(p_W) \geq \deg(p_V) - \deg(p_{V/W}) = \dim(V) - \dim(V/W) = \dim(W)$$
which, again by the fact, tells us that $\deg(p_W) = \dim(W)$, and so the restriction of $T$ to $W$ is also cyclic.