kaplanelad/shellfirm

Terminal guardrails for humans and AI agents — intercepts dangerous commands with context-aware challenges before the damage is done.