Guardrails for offensive AI agents and why instructions are never enoughdest.host3 points·snorbleck··0 commentsOpen articleSaveView on HN