[guided]The point of this step is to turn pointwise curvature information into constants that work everywhere on $X$. The fixed bundle $F$ may have curvature with negative directions, but compactness ensures that this negativity is uniformly bounded.
We first use the Kähler metric $\omega$ and the Hermitian metric $h_F$ to measure tensors in $T^{1,0}X\otimes F$. Define
\begin{align*}
S:=\{(x,u)\mid x\in X,\ u\in T_x^{1,0}X\otimes F_x,\ |u|_{\omega,h_F}=1\}.
\end{align*}
For a point $x\in X$ and a tensor
\begin{align*}
u=\sum_{j,\alpha}u_{j\alpha}\,\partial_{z_j}\otimes e_\alpha\in T_x^{1,0}X\otimes F_x,
\end{align*}
written in a unitary frame, define
\begin{align*}
B_F(x,u):=\sum_{j,k,\alpha,\beta} i\Theta(F,h_F)_{j\bar k\alpha\bar\beta}(x)\,u_{j\alpha}\,\overline{u_{k\beta}}.
\end{align*}
Although this formula uses a frame, the value does not depend on the chosen unitary frame: it is precisely the curvature Hermitian form evaluated on the tensor $u$.
The function $B_F:S\to\mathbb R$ is continuous because the Chern curvature tensor of the smooth Hermitian metric $h_F$ has smooth coefficients. The space $S$ is compact because it is the unit sphere bundle of the finite-rank Hermitian vector bundle $T^{1,0}X\otimes F$ over the compact base $X$. Therefore $B_F$ attains a minimum. Define
\begin{align*}
A:=\max\{0,-\min_S B_F\}.
\end{align*}
This choice gives $B_F(x,u)\ge -A$ whenever $|u|_{\omega,h_F}=1$. For an arbitrary nonzero $u$, apply this inequality to $u/|u|_{\omega,h_F}$ and multiply by $|u|_{\omega,h_F}^2$; for $u=0$ the inequality is immediate. Thus, for every $x\in X$ and every $u\in T_x^{1,0}X\otimes F_x$,
\begin{align*}
B_F(x,u)\ge -A|u|_{\omega,h_F}^2.
\end{align*}
Now we do the same for the positive curvature of $L$. Define
\begin{align*}
S_X:=\{(x,\xi)\mid x\in X,\ \xi\in T_x^{1,0}X,\ |\xi|_\omega=1\}.
\end{align*}
Since $\omega_L=i\Theta(L,h_L)$ is positive, $\omega_L(\xi,\bar\xi)>0$ for every nonzero $\xi\in T_x^{1,0}X$. The function $(x,\xi)\mapsto \omega_L(\xi,\bar\xi)$ is continuous on the compact set $S_X$, so it has a minimum. That minimum is strictly positive because the function is positive at every point of $S_X$. Set
\begin{align*}
a:=\min_{S_X}\omega_L(\xi,\bar\xi)>0.
\end{align*}
Scaling again gives, for every $x\in X$ and every $\xi\in T_x^{1,0}X$,
\begin{align*}
\omega_L(\xi,\bar\xi)\ge a|\xi|_\omega^2.
\end{align*}[/guided]