Permission Modes

Permission Modes

Permission modes decide how much the XHack Agent is allowed to do without asking. You set the mode per chat, so a careful engagement and a fast internal test can each have their own level of autonomy. It is one click in the status bar.

Switch the mode

1. Look at the bottom status bar for the mode chip. It shows the current mode.
2. Click the chip to cycle to the next mode, or press Shift+Tab.

The modes

• Manual, the Agent asks before every action. Best for production targets or anything sensitive.
• Allow Edits, file edits are auto-approved, but other actions still ask. A good middle ground when the Agent is mostly writing notes, code, or reports.
• Auto. Everything in this chat is auto-approved. Use it on authorized targets when you want the Agent to move without stopping to ask.
• Plan, read-only. The Agent investigates and proposes an approach but changes nothing. Great for a first pass on an unfamiliar target.

Approve or deny an action

In Manual (and for actions Allow Edits does not cover), a card pops up centered over the chat with three buttons:

• Allow (key Y), permit this one action.
• Always Allow (key A), permit this and stop asking for the same kind of action.
• Deny (key N), refuse and let the Agent try something else.

Guardrails never switch off

Even in Auto, and in the fully hands-off ⚡ Auto autonomous run (in the status bar, or Ctrl+Shift+A), destructive commands are hard-blocked. Actions that could wipe data or damage a system are refused no matter what mode you are in, so a fast setting can never turn into an accident.

Related

• Sub-Agents. Each agent runs under a permission mode
• Autonomous Agent, the fully hands-off run
• Restrictions & Prohibited Activities, what is always off-limits