claude-switch-models-setup

SUSPICIOUS: the core purpose is legitimate and most scanner hits are documentation/context, but the skill's footprint is broader than simple profile switching. The main concern is its normalization of `--dangerously-skip-permissions`, plus transitive skill installation and extensive direct manipulation of sensitive Claude config/state across profiles. No clear evidence of credential exfiltration or malware, but the trust and permission model is too aggressive for the stated task.