I can confirm this problem. Here are some details for Thomas to fix this bug:

I think the problem is, that the "skilled" talent is handled twice. First in talents.c (ModifyByTalent(), lines 561-577) and a second time in init.c (InitializeGame(), lines 1664-1681).

Each routine calls ModifySkillBy() with different values for the parameter "no_magic", which leads to an "Illegal skill advancement" error, when the this function is called from init.c (no_magic = TRUE).

It always has, ever since it was introduced.. TB claimed he fixed it from v1.1.0 to 1.1.1 but that wasn't accurate, as it still occurred.

Perhaps in the 1.2.x versions this will finally be addressed, but I think (just from inferences) that TB is working hard on JADE so might not tackle ADOM problems right now.

Who would ever take skilled anyway? =)

I agree. It's a waste of a talent. If it gave you some extra(2-3) skills, maybe, but +2 to each skill? That's nothing.

According to sorear, "Skill increases come in two flavors, ones that also increase potential and ones that don't; Skilled uses the wrong type and the game asserts after potential goes negative. "

ModifySkillBy() is now called with "no_magic=FALSE".

According the r.g.r.a bug list, the -foo- Skilled talents also do twice what it says in the manual - is that also fixed?

I could not reproduce this problem. The values in the manual and source code match, you gain "+2" for selecting "Skilled" and "+3" for selecting "Very skilled".