Ragnar Groot Koerkamp

yeah `bt validate` does this so we're good. I don't do it while generating for simplicity I suppose.

In that case, `p` (or `o/e/x`) should probably be passed to `WF_NEXT` as well, since currently algorithm 1 does not actually use `p`.

I think stress testing the input and output validators sense, and this is a reasonable way to do it. One thought: in NWERC we always use some library (header file),...

Right yes, updating the contest `skel` dir before adding problems is a decent workaround for sure.

This is pretty nice! I see the current skel is already quite verbose anyway also re. generators so having a bit more for invalid is fine. And an explicit example...

As an example, when aligning `AC` and `AACC` with cost model `(sub=1, open=1, extend=1)`, triple_accel correctly returns cost `3`, but the cigar is `X=X=` instead of `=XX=`.

It's quite unfortunate this project doesn't have a license even though it has so many stars. Could you indeed add a license so that it's clear whether it's fine to...

@Ashe Note that you could already just take the updated `grimshot` file and put it somewhere in your path (at least that's what I did).

@fluix-dev Oh, I didn't know of `slurp -o` before. That simplifies things since now we can use the same `jq` command as the existing `window` option, and just do `slurp...

`selection` could maybe change to e.g. `anything`