I echo Hillel on this: while not perfect, the Wang and Sternberg STOP-IN approach - with a 47 bp insertion with a primer landing site - is more economical in reagents and time. If you consider p = probability of a cut at the site of your provided guide RNA, then two cuts in the same chromosome with be p^2. In the two-site gene deletion, if p for each site is 0.10, then the p of two cutting is 0.01. For a STOP-IN it is simply p = 0.1, and so much more efficient: 10% vs. 1%. To unpack that, to be ~99% confident of getting a mutation, for one guide in the above example you’d need to genotype ~40 Rol/dpy-10(cn64) co-CRISPR F1s. For two guides you’d need ~400 animals. So it matters.
Additionally, with STOP-IN one can “shop” for the optimal guide RNA. We usually do this in the first half of the gene, in exons, and biasing toward the 5’ end of the gene. (Remember, you’d like to truncate the resulting protein earlier rather than later, plus Nonsense-Mediated Decay does not work well in the last two exons of a gene. I also avoid locating a guide right next to introns. I don’t know if this avoid use of cryptic splicing, but presumably location in the middle of an exon would minimize that.)
We use a combination of algorithms and chart a middle course: SFU as noted by Hillel, CRISPOR (now at UC Santa Cruz) and WU-CRISPOR. We also use a general principle of shooting for a guide with a GG dinucleotide at positions -1/-2,a s originally predicted by Wang/Sabatini/Lander 2014, Fig 3F. By that, the “ideal” guide is a “GCGG” and the “worst” guide is a “UNUU.” The desirability of a GG has been validated in elegans by Farboud Meyer 2015. With STOP-IN you can shop for these. Our anecdotal experience is that guides based on sequence/algorithms have been much more efficient on average than guides based on location.
Also, we never work with guides with low specificity, e.g. we are usually in the 97-100 range but never less than low 90s using the MIT algorithm. We check predicted off-site targets but that is rarely an issue. For those of us used to dirty old mutagenized backgrounds and familiar with the rate of accumulation of mutations in cultured elegans, CRISPR is a dream of precision. Remember 20-30% of all edits will have misrepairs, but you don’t care with STOP-IN.
For all insertions/deletions, CRISPR, KO consortium or whatever, we use triplex PCR. One primer inside, two flanking. As Hillel notes about sizes: bands compete and not always as expected. Keep your products small (always >500 bp, preferably smaller) and not too different in size from each other. Also, we are VERY fussy about primer design/specificity and use high Tm of 60. We pilot them before the edit but of course you do not have a positive control because you have not yet edited it. For triplex, whenever possible we use three controls single worm lysates: +/+, m/+ and m/m. You would be surprised at how often the products interact in strange ways. For autosomal m/+ simply cross into your mutant and pick het male progeny for lysis. LGX requires another generation and a Dpy or Unc.