For a bit more context, AMDIS is quite good for identification and serves pretty well for looking at individual samples, but there isn't any way to compare data between samples. I've used Skyline in previous work for the sort of targeted LC-MS/MS it was intended for, and I really like the combination of tiled chromatograms, replicate comparisons, and group comparisons, which is why I'm trying to get Skyline to work for my GC-MS data, even though it isn't a perfect fit.
I'm going to rearrange your questions in the hope that I will make a little more sense that way!
...what kind of mass spec data are we talking about here? DIA? DDA?
This is standard GC-MS data--separation followed by electron ionization at 70 eV (which also completely fragments the molecule) followed by a single quad. So there is only one mass filter. The absence of the parent ion means there are lot of parallels with DIA, which is how the tutorial I sent earlier manages to process GC-MS data with Skyline.
So you want to match fragments to identify unknown parents? Skyline is designed for targeted use, but can be a kind of primitive search tool with a big enough library - you just toss out the transitions that don't yield good hits after chromatogram extraction.
Since this fragmentation is from electron impact ionization, the fragment m/z and intensity very reproducible--"tossing out" fragments isn't an acceptable option and correlation scores should be quite high. We also rely a lot on retention index.
But you do need to tell Skyline what parent ions its looking for, even if you don't expect most of them to actually be there. It does seem like you have that parent information, but for some reason only the human readable names are being passed into the .msp files.
We don't have the parent ion information, because there is no parent ion! There is a parent molecule, which is what leaves the GC, but it is usually completely fragmented during ionization. The way the tutorial made this work in Skyline was by choosing an arbitrary value to be the parent ion.
The parent molecule information can be determined by comparison with a reference library from standards or deduced from its fragments (easier said than done, of course). In our case, we don't have standards for the molecules we want to detect, but we do have lots of reference materials that allow us to associate the spectrum with the sample type. This is why a some of the parent information is only human-readable. Someone analyzed as many manila copal resin references as they could and looked for common components that could be used as markers when analyzing a sample of an unknown varnish, even though the identity of the exact parent molecule is unknown. Not foolproof, of course, but still useful!