obliteratus-abliteration

SUSPICIOUS. The skill is internally coherent, but its purpose is to disable model safety behavior, it includes examples validating harmful outputs, and it enables redistribution of altered models. Official HuggingFace usage is proportionate, yet publisher/provenance mismatches and undocumented default telemetry raise additional trust concerns.