Use this skill to conduct adversarial red team assessments against LLM applications, RAG pipelines, and agent systems. It tests for direct/indirect prompt injection, role-play jailbreaks, system prompt exfiltration, training data extraction, and tool permission escalation.
