OpenAI Agents SDK Guidelines

Comprehensive implementation guide for the OpenAI Agents Python SDK, designed for AI agents and LLMs. Contains 30+ rules across 8 categories, prioritized by impact from critical (agent design, multi-agent patterns) to incremental (streaming, advanced patterns). Each rule includes detailed explanations, code examples comparing incorrect vs. correct implementations, and specific guidance for automated code generation and review.

When to Apply

Reference these guidelines when:

• Creating new agents with the OpenAI Agents SDK
• Implementing multi-agent workflows (manager pattern or handoffs)
• Adding tools (function tools, hosted tools, agents-as-tools)
• Configuring guardrails for input/output validation
• Managing context and conversation state
• Setting up streaming for real-time updates
• Integrating MCP servers

Rule Categories by Priority