This is the guide provided by OpenAI to help use agents SDK. Several key points cited in the following.

First, write good prompts to configure, for example use advanced models such as o3-mini to write ““You are an expert in writing instructions for an LLM agent. Convert the following help center document into a clear set of instructions, written in a numbered list. The document will be a policy followed by an LLM. Ensure that there is no ambiguity, and that the instructions are written as directions for an agent. The help center document to convert is the following {{help_center_do}}”.

While it’s tempting to immediately build a fully autonomous agent with complex architecture, customers typically achieve greater success with an incremental approach. Orchestration should be considered. But if a single agent can do the work, use single agent, note an effective strategy for managing complexity without switching to a multi-agent framework is to use prompt templates. Rather than maintaining numerous individual prompts for distinct use cases, use a single flexible base prompt that accepts policy variables. for example “”” You are a call center agent. You are interacting with {{user_first_name}} who has been a member for {{user_tenure}}. The user’s most common complains are about {{user_complaint_categories}}. Greet the user, thank them for being a loyal customer, and answer any questions the user may have!”””

Multi-agents has two broadly applicable categories: Manager pattern and decentralized pattern. Manager pattern is more or so like parent-child agents system versus decentralized one.

In the diagram below, we combine LLM-based guardrails, rules-based guardrails such as regex,
and the OpenAI moderation API to vet our user inputs.

Lastly, building guardrails is important, the following heuristic to be effective: Focus on data privacy and content safety, optimize both security and user experience. The Agents SDK treats guardrails as first-class concepts, relying on optimistic execution by default. Under this approach, the primary agent proactively generates outputs while guardrails run concurrently, triggering exceptions if constraints are breached. Human intervention is a critical safeguard enabling you to improve an agent’s real-world
performance without compromising user experience. It’s especially important early in deployment, helping identify failures, uncover edge cases, and establish a robust evaluation cycle.