Integration
Safety Confirmation
To ensure that all actions initiated by the Agent that might affect your local system are under your full control, Z Code enables a Manual Confirmation Mechanism by default.
This mechanism is a core part of our security strategy, giving you the Final Veto Power over Agent behavior to prevent accidental or unauthorized operations.
Safety Tip: Always carefully read the content about to be executed before granting permission.
Workflow
- Trigger Confirmation: When the Agent attempts any potentially risky action (such as running command-line scripts, accessing network resources, or modifying critical files), its execution flow pauses automatically.
- Pop-up Request: A clear confirmation request appears on the interface, containing the specific command or action the Agent plans to execute.
- User Decision: You must manually confirm the request for the Agent to proceed.
Decision Options
| Option | Description | Recommended Scenario |
|---|---|---|
| Allow | Authorizes the Agent to execute only this specific action. | Temporary or uncertain single tasks. |
| Reject | Prohibits the Agent from executing the current action. The Agent will abort the current task. | When instructions don't match expectations or represent a security risk. |
| Always Allow | Authorizes the Agent to execute the current and all future actions of the same type in this session without asking again. | When the operation is trusted and needs to be executed frequently (e.g., routine compilation). |
Note: "Always Allow" temporarily upgrades Agent permissions. Please use it only when you are certain the operation is safe and trustworthy.
Applicable Scenarios
- Running Third-Party Scripts: When the Agent attempts to run Python, Shell, or Node.js scripts in your project.
- Network Requests: When the Agent needs to access external APIs via
curlor other methods. - System-Level Commands: When executing commands that may modify system configuration or delete files.