The big challenge at the moment is coverage; getting enough of the cells successfully edited by the gene therapy to make a meaningful difference. There are promising signs in mice, such as the muscular dystrophy study from earlier this year [1], but in general this is still an ongoing battle, without a standard approach.
The coverage issue is why you can't go out and get any one of a hundred edits [2] and expect things to happen as a result. As soon as that is robustly solved, well, gene therapies will be out there in the medical tourism market, starting with telomerase, myostatin, and follistatin, I'd imagine. Most likely the second of those up front since its already fairly robustly proven in humans via antibody therapies. [3]
I would be extraordinarily surprised to find that there were no humans already tinkering with their myostatin balance via genes or antibodies, given the number of people with access to the technology, and the ease with which such a therapy could be assembled.
The biggest challenge with gene editing is knowing what to edit! If you're trying to correct a single base-pair genetic defect, it's more straightforward, but even for something as "simple" as height, we have little insight into what genes control that characteristic nor how to actually modify them.
It's nice there are more gene editing methods but this just moves the problem back where it always belonged: delivery of the vector to the cells that matter, and knowing what edit to make are still extremely hard problems. The latter seems to be the ultimate challenge: our current understanding of genomics is still very much stuck in the "the function of a gene is assumed to be the largest eigenvalue of its principle component analysis". In complex, redundant systems with feedback (like the genome and its associated cellular machinery) it's basically impossible to fix an actual problem, with limited side effects, by just modifying a single gene.
Recognizing one human among seven billion by their DNA would probably take an unwieldy large amount of recognition sequence. Releasing a weapon with a killing payload and letting it be subject to natural selection risks it evolving to kill more people, or all people. In general, bioweapons are scary because they can't by their nature have reliable off switches.
They are also a heap of excessive effort when a sniper rifle or drone missile works fine.
Actually, genome editing would rather provide a defense against such an attack as you could specifically augment your DNA to look like someone else or be heterogeneous (within limits of autoimmune responses).
I also fail to see how it could help to create something like FOXDIE (I have only read a short wiki article about it)? I mean you can create viruses with a certain sequence already now, so is no clear connection between creating a targeted virus and genome editing? Sure, you can edit the genome but injecting a single malicious gene is much easier (which seems to be the case for the fictional FOXDIE)