CoP: Agentic Red-teaming for Large Language Models using Composition of Principles | Read Paper on Bytez