Is GPT-6 Sol just completely skill-lazy for anyone else?

Users are reporting that GPT-6 Sol is failing to properly utilize configured skills, a problem not observed with GPT-5.6. Despite explicit instructions and correct configurations, GPT-6 Sol reportedly ignores relevant skills, pulling data from random sources or failing on basic requests like weather queries. This behavior forces users to constantly monitor and re-instruct the model, undermining its intended purpose and leading to frustration among the developer community.