LLM-first CLI and MCP toolkit that lets humans and agentic coding tools collaborate on application security testing