Permission Mode

You might say

Routine changes inside this project can continue automatically, but ask before deletion, wider access, or external actions. Which permission setting fits?

A set of control presets defining file system boundaries, network access limits, and user confirmation triggers for an AI agent.For example, one Codex permission option can allow routine reads, writes, and local commands inside the current workspace, then ask before network access or crossing that boundary. Names and combinations differ across tools. Check the sandbox scope, when prompts appear, and who reviews them before choosing.

Codex permission modes

Quick checkChoose the best answer

Which statement about permission modes is correct?

Plan / read-only is mainly for reading and analysis without changing code; after confirming the plan, move to the execution phaseFull access only changes model capability and does not affect the AI's access to files or network resourcesAsk for approval means the AI can never change code and may only ask the user questions

You can say this to an AI Agent

Before any code development, first confirm the requirements with me, read the necessary context in Plan or read-only mode, and prepare a clear approach and plan. Do not change code or perform other development actions until I confirm the plan. After confirmation, choose the permission settings that match the task's access scope; explain the impact before widening the scope, then inspect the Diff and run the tests.

Learn next

1Sandbox→2Project Rules→3API Key→

Recommended tool

Web fine-tuning tool · Selector Pro

Select page elements for precise AI edits

Click a button, text, or card, note the change, then copy its exact location to AI.

Open Selector Pro website