A demonstration showcases the integration of multiple AI agents using Google's Agent Development Kit (ADK) and the Agent-to-Agent (A2A) protocol, highlighting a novel approach where a language model (LLM) orchestrates tasks dynamically without hardcoded workflows. The system features three specialized agents—Research, Writer, and Security—that do not have prior knowledge of each other, with the orchestrator discovering them at runtime via a registry service. The A2A protocol allows AI agents to communicate and coordinate tasks securely, promoting interoperability across different platforms. Unlike traditional rigid orchestration, the LLM decides the sequence of agent activation and task completion, showcasing flexibility and efficiency. In this setup, the Security Agent uses GitGuardian to scan content for exposed secrets, illustrating a critical aspect of security in AI-generated content. The demo emphasizes the value of dynamic discovery, model-agnostic design, and the necessity of implementing guardrails such as turn limits to prevent infinite loops, while also demonstrating real-time security scanning capabilities.