How does it handle agent errors or infinite loops?

The skill includes documented workarounds for 9+ common errors, including MaxTurnsExceededError, ToolCallError, and Zod schema type mismatches.

Does this skill support multi-agent handoffs?

Yes, it provides specific implementation patterns for delegating tasks between specialist agents using handoff descriptions and triage logic.

Is it compatible with Cloudflare Workers?

Yes, it includes experimental patterns and limitation guidance for deploying text-based agents to Cloudflare's serverless environment.

How does it ensure structured data output?

It utilizes Zod schemas for both tool parameters and final agent responses to ensure type-safe, predictable, and validated AI outputs.

Can I use this for voice-based AI applications?

Absolutely, it includes templates for Realtime voice agents using WebRTC/WebSocket and best practices for secure session token management.

OpenAI Agents SDK Integration

Name: OpenAI Agents SDK Integration
Author: jezweb

byjezweb

•

117

•

数据科学与机器学习

Build and deploy sophisticated AI agent workflows featuring multi-agent handoffs, realtime voice, and type-safe tool execution.

This skill equips Claude to architect and implement complex agentic systems using the OpenAI Agents SDK. It provides specialized guidance for creating both text and realtime voice agents, managing multi-agent handoffs between domain specialists, and enforcing strict type-safety through Zod schemas. By incorporating patterns for human-in-the-loop (HITL) approvals, input/output guardrails, and production deployment to platforms like Cloudflare Workers and Next.js, it streamlines the development of robust, reliable AI applications while proactively preventing common SDK errors and orchestration failures.

主要功能

01Realtime voice agent setup using WebRTC and secure server-side session tokens.

02Built-in human-in-the-loop (HITL) patterns for safety-critical tool approvals.

03Multi-agent orchestration with autonomous or code-based delegation patterns.

04Type-safe tool definitions and structured outputs using Zod schema validation.

05Comprehensive error handling for common issues like infinite loops and schema mismatches.

06117 GitHub stars

使用场景

01Building multi-agent triage systems that route users between billing, tech, and sales specialists.

02Implementing secure, serverless AI agents on Cloudflare Workers with automated input/output guardrails.

03Developing low-latency voice assistants with handoff capabilities and browser-based WebRTC sessions.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jezweb/claude-skills openai-agents

For use in Claude.ai and ChatGPT

Download Skill