
openai-agents
by MattMagg
Claude Code plugins for building AI agents across frameworks (Google ADK, OpenAI, and more)
SKILL.md
name: OpenAI Agents description: Workflow patterns and gotchas for OpenAI Agents SDK. Directs to RAG for implementation.
OpenAI Agents Workflow
When to Choose OpenAI Agents
- Already using OpenAI API and models
- Need handoff patterns between specialized agents
- Want function calling with type safety
- Building conversational assistants
Decision Framework
Architecture Patterns
| Need | Pattern | RAG Query |
|---|---|---|
| Single specialist | Basic Agent | "Agent basic example" |
| Route to specialists | Handoffs | "agent handoff routing" |
| Sequential steps | Agent chains | "agent chain pattern" |
| Persistent context | Memory | "agent memory conversation" |
| Background processing | Async runner | "async agent runner" |
Query RAG: mcp__agentic-rag__query_sdk("pattern name", sdk="openai", mode="build")
Critical Gotchas
These cause silent failures or confusion:
- Use
@tooldecorator - Functions without it won't be recognized - Type hints are required - Parameters need type annotations for schema generation
- Docstrings define the schema - LLM uses docstring for understanding, not just comments
- OPENAI_API_KEY env var - Must be set; SDK doesn't prompt for it
- Model names are specific -
gpt-5.2,gpt-5.2-codex(notgpt5orGPT-5.2) - Handoff descriptions drive routing - Vague = poor decisions
Workflow: Creating an OpenAI Agent
Step 1: Environment Setup
RAG Query: mcp__agentic-rag__query_sdk("openai agents installation", sdk="openai", mode="explain")
Step 2: Basic Agent
RAG Query: mcp__agentic-rag__query_sdk("Agent class definition", sdk="openai", mode="build")
Step 3: Tool Definition
RAG Query: mcp__agentic-rag__query_sdk("@tool decorator example", sdk="openai", mode="build")
Type hints + docstring = working tool.
Step 4: Running the Agent
RAG Query: mcp__agentic-rag__query_sdk("Runner execution", sdk="openai", mode="build")
Step 5: Handoffs (if multi-agent)
RAG Query: mcp__agentic-rag__query_sdk("handoff between agents", sdk="openai", mode="build")
Step 6: Memory (if conversational)
RAG Query: mcp__agentic-rag__query_sdk("agent memory persistence", sdk="openai", mode="build")
Common Error Patterns
| Symptom | Likely Cause | RAG Query |
|---|---|---|
| Tool not found | Missing @tool decorator | "tool decorator usage" |
| Schema error | Missing type hints | "tool type annotations" |
| API error | Invalid API key | "openai authentication" |
| Model error | Wrong model name | "supported models" |
| Handoff fails | Bad description | "handoff description" |
Advanced Features
Query RAG when you need:
- Streaming responses:
"streaming agent output" - Parallel tool calls:
"parallel tool execution" - Custom runners:
"custom Runner implementation" - Guardrails:
"agent guardrails safety" - Tracing:
"agent tracing debugging"
Score
Total Score
Based on repository quality metrics
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
Reviews
Reviews coming soon