[guided]The complex structure on $X$ decomposes complex-valued $1$-forms by type. Let $\mathcal{A}^{1,0}(X)$ be the smooth $(1,0)$-forms and let $\mathcal{A}^{0,1}(X)$ be the smooth $(0,1)$-forms. For smooth $(p,q)$-forms, define the operators
\begin{align*}
\partial: \mathcal{A}^{p,q}(X) &\to \mathcal{A}^{p+1,q}(X), \\
\bar{\partial}: \mathcal{A}^{p,q}(X) &\to \mathcal{A}^{p,q+1}(X)
\end{align*}
by the type decomposition $d\beta = \partial\beta + \bar{\partial}\beta$ for each $\beta \in \mathcal{A}^{p,q}(X)$. Let
\begin{align*}
*_h: \mathcal{A}^1(X;\mathbb{C}) &\to \mathcal{A}^1(X;\mathbb{C})
\end{align*}
be the Hodge star operator determined by $h$; on $1$-forms the corresponding codifferential is $d_h^* = -*_h d *_h$.
For every smooth complex-valued $1$-form $\alpha \in \mathcal{A}^1(X;\mathbb{C})$, there are unique forms $\alpha^{1,0} \in \mathcal{A}^{1,0}(X)$ and $\alpha^{0,1} \in \mathcal{A}^{0,1}(X)$ such that
\begin{align*}
\alpha = \alpha^{1,0} + \alpha^{0,1}.
\end{align*}
Now assume $\alpha$ is harmonic. In particular, $d\alpha = 0$ and $d_h^*\alpha=0$. Define the two $(1,1)$-forms
\begin{align*}
A := \bar{\partial}\alpha^{1,0} \in \mathcal{A}^{1,1}(X),
\qquad
B := \partial\alpha^{0,1} \in \mathcal{A}^{1,1}(X).
\end{align*}
Since $X$ has complex dimension $1$, there are no nonzero smooth forms of type $(2,0)$ or $(0,2)$. Thus
\begin{align*}
\partial\alpha^{1,0}=0,
\qquad
\bar{\partial}\alpha^{0,1}=0.
\end{align*}
Decomposing $d\alpha=0$ by type therefore gives
\begin{align*}
0=d\alpha=A+B.
\end{align*}
This equation alone is not enough to force $A$ and $B$ to vanish separately, because both have type $(1,1)$. The missing information is the co-closed condition. Since $d_h^*\alpha=-*_h d *_h\alpha=0$ and $*_h$ is an isomorphism on $1$-forms, we obtain
\begin{align*}
d(*_h\alpha)=0.
\end{align*}
The Hodge star satisfies $*_h\alpha^{1,0}=-i\alpha^{1,0}$ and $*_h\alpha^{0,1}=i\alpha^{0,1}$, so
\begin{align*}
*_h\alpha=-i\alpha^{1,0}+i\alpha^{0,1}.
\end{align*}
Using again that the $(2,0)$ and $(0,2)$ terms vanish in complex dimension $1$, the equation $d(*_h\alpha)=0$ becomes
\begin{align*}
0=d(*_h\alpha)=-iA+iB.
\end{align*}
The system
\begin{align*}
A+B=0,
\qquad
-iA+iB=0
\end{align*}
implies $A=0$ and $B=0$. Hence
\begin{align*}
\bar{\partial}\alpha^{1,0}=0,
\qquad
\partial\alpha^{0,1}=0.
\end{align*}
The first equation says exactly that $\alpha^{1,0}$ is a holomorphic differential. The second says that $\alpha^{0,1}$ is anti-holomorphic, equivalently that $\alpha^{0,1} = \overline{\omega}$ for a unique holomorphic differential $\omega \in H^0(X,\Omega^1)$.
Conversely, let $\omega \in H^0(X,\Omega^1)$. Then $\omega$ is a smooth $(1,0)$-form satisfying $\bar{\partial}\omega = 0$. Since $X$ has complex dimension $1$, the form $\partial\omega$ has type $(2,0)$ and must vanish. Hence
\begin{align*}
d\omega = \partial\omega + \bar{\partial}\omega = 0.
\end{align*}
For a Hermitian metric on a Riemann surface, the Hodge star acts on $(1,0)$-forms by $*_h\omega = -i\omega$. Therefore
\begin{align*}
d(*_h\omega) = d(-i\omega) = -i\,d\omega = 0.
\end{align*}
The codifferential on $1$-forms is $d_h^* = -*_h d *_h$, so the last equality gives
\begin{align*}
d_h^*\omega = -*_h d(*_h\omega) = 0.
\end{align*}
Together with $d\omega = 0$, this proves that $\omega$ is harmonic. Complex conjugation gives the same conclusion for every $\overline{\omega} \in \overline{H^0(X,\Omega^1)}$. This proves
\begin{align*}
\mathcal{H}^1_h(X;\mathbb{C})
=
H^0(X,\Omega^1)
\oplus
\overline{H^0(X,\Omega^1)}.
\end{align*}
The sum is direct because a form of type $(1,0)$ and a form of type $(0,1)$ can be equal only if both are zero.[/guided]