I want to introduce the sequence for a tetracysteine tag in a protein of interest, in vivo. The nucleotide sequence is just 18bp long. I know I can use Crispr to knock-in genes but most papers are dealing with genes of larger sizes.
Smaller insertions are more efficient. You may use smaller homology arms. For a detailled study see: Article Efficient CRISPR-Cas9-Mediated Knock-In of Composite Tags in...
I put together a few references that go over knock-in methodologies, especially around small (ssODN-mediated) edits here: https://invivobiosystems.com/crispr-model-creation/zebrafish-genome-editing/zebrafish-precise-sdm/#key-references
As Uwe mentioned, smaller knock-ins (though larger knock-ins like fluorescent tags are increasingly being done; I highly recommend giving this a read: Article Efficient targeted integration directed by short homology in...
) are typically more efficient than larger insertions, especially when it comes to HDR activity.
That said, there are a lot of publications out there now touting increased HDR efficiency with different flavors of repair template, but few of them validate this effect with more than one or two loci. I would also highly recommend this study (also linked in the InVivo page) that looks at the effects of a myriad of ODN repair template designs at several loci, and also delves into proposed repair mechanisms underlying complex repairs associated with ssODN-mediated knock-in: https://dmm.biologists.org/content/11/10/dmm035352