Permissions

• Key points are in the visual diagram above.

A coding agent without permissions is a remote shell as the model.

OS permissions: the model is never root by default.

Scenario

OS permissions: the model is never root by default.

Solution

In Claude Agent SDK, apply Permissions to this scenario: OS permissions: the model is never root by default. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Permissions (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Permissions happens in the code.

1# Permissions — minimal example2from openai import OpenAI3
4client = OpenAI()  # create API client5
6# Ask the model to explain this topic7response = client.chat.completions.create(  # core API call for Permissions8    model="gpt-4o-mini",9    messages=[10        {"role": "system", "content": "You explain permissions clearly."},11        {"role": "user", "content": f"What is permissions?"},12    ],13    temperature=0,14)15print(response.choices[0].message.content)  # show output for debugging

• Skipping evaluation for Permissions before production
• No logging or tracing around claude permissions steps
• Ignoring cost and latency implications

Quick recap — the most important points from this module.