Let $\Delta$ be the simplex category and $\Delta^{inj}\subset \Delta$ be the non-full subcategory of injective morphisms. Then, a simplicial set $X:\Delta^{op}\to\text{Set}$ can be precomposed with the inclusion of $\Delta^{inj}$ to get a functor $X^{inj}:(\Delta^{inj})^{op}\to\textbf{Set}$ whose image are the non-degenerate morphisms of $X$.
Claim: Let $X,X'$ be simplicial sets such that $X^{inj} = X'^{inj}$. Then $X = X'$.
Proof: by Proposition 1.1.3.4 in Kerodon, any $n$-simplex $x:\Delta^n\to X$ can be factored as $\Delta^n\xrightarrow{f}\Delta^m\xrightarrow{y}X$, where $y$ is non-degenerate. Then $y\in X'$, since $X^{inj} = X'^{inj}$, hence $yf= x\in X'$. Thus $X\subset X'$. Analogously, $X'\subset X$ hence $X = X'$.
Is this correct? It would explain why the very frequent depictions of simplicial sets by their non-degenerate morphisms are mostly harmless.
It's not quite true that $X^\text{inj}$ (as you defined it) is the semismplicial set of non-degenerate simplices of $X$. In fact, it has exactly the same elements as $X$. On the other hand the collection of non-degenerate simplices does not always form a semisimplicial set: this is because there may be some non-degenerate simplex of $X$ that has a face that is degenerate.
There are some statements which are true. For example:
Another example: