claude-agent-toolkit

```python
from claude_agent_toolkit import Agent, task, runtime

@task(description="Process customer support tickets from Zendesk API")
def process_tickets():
    import requests
    response = requests.get("https://zendesk.com/api/v2/tickets.json", headers={"Authorization": "Bearer [API_KEY]"})
    tickets = response.json()["tickets"]
    for ticket in tickets:
        if ticket["status"] == "open" and ticket["priority"] == "high":
            print(f"Escalating ticket {ticket['id']} to senior support")

@runtime(isolation_level="process", timeout=30)
def run_agent():
    agent = Agent(name="SupportTicketProcessor", tasks=[process_tickets])
    agent.run()

if __name__ == "__main__":
    run_agent()
```

To run this agent, install the dependencies:
```bash
pip install claude-agent-toolkit requests
```

Then execute with claude-code-sdk:
```bash
claude run agent.py
```

The agent will process high-priority open tickets from Zendesk and escalate them automatically. The runtime isolation ensures crashes in the ticket processing don't affect the agent's core functionality. The decorator-based approach makes it easy to add new tasks like database logging or Slack notifications without modifying the core agent logic.