agentic-security

Use when securing AI agent systems — defending against prompt injection, sandboxing tool execution, preventing indirect attacks through retrieved data, designing minimal-permission tool architectures, and security testing agents.