You can register scripts, commands, and AI prompts as actions in the **Agentic Assistant Hub**. After a script, command, or AI prompt is registered as an action, it can be added to one or more agents. The agents can then execute the action as part of plans.
To register actions, ensure you have the correct permissions. For more information, see [Agentic Assistant role-based access control](../agentic-assistant-role-based-access-control).
When you register an action, you provide a description, goal, and, optionally, a few-shot examples. This information helps agents understand how the action should be used.
When registering or editing an action, you can choose which specific inputs and outputs are visible to the LLM. For example, a script might have two inputs and five outputs, but for this action, only one input and two outputs are required, and only those are included in the action. This helps to create more focused actions and reduces unnecessary complexity.
A single content item can be registered as different actions, with each action using different inputs and outputs from the same script. Only register the same script, command, or AI prompt as a new action if it is required for your use case, as providing an agent with many actions with overlapping abilities can reduce the ability of the agent to choose the most appropriate action.
While you can create multiple actions from a single content item, each action must have a different name.
If you try to register a script, command, or AI prompt that is already registered, you are presented with a list of the actions already using it, and you can review and decide if any of them are relevant for your current use case. If not, you can register the script, command, or AI prompt again as a different action.
How to register an action
Do one of the following:
* Click the **Agentic Assistant** icon in the upper right hand corner and expand the side panel  to access the **Agentic Assistant Hub** menu item. From the **Actions** tab of the **Agentic Assistant Hub**, click **Register new action**.
* Within the script creation or editing screen, click <img src="/docs/images/6d89b981f06a2e4a.png" alt="three-dots.png" data-size="line"> when viewing or editing a script and select **Register as action**.
* Within the **AI Prompts** library, select a prompt and click the more options icon to **Register as action**.
If you clicked **Register as action** from the **Scripts** or **AI Prompts** page, the name is prepopulated in the **Content chosen** field. If you clicked **Register action** from the **Agentic Assistant Hub**, select script, command, or AI prompt for the **Type of content** and select the content you want to register.
Enter an action **Name**, a short description of what the action does. Example: `Extract email`.
Describe the **Goal** of the action.
(Optional) By default, **Mark action as sensitive to require user approval** is selected, and the agent prompts the user to approve before executing the action. If you do not want the action marked as sensitive, clear the checkbox.
(Optional) Provide **Few-shot examples** to help the agent understand the context and the appropriate situations to invoke a specific action.
Click **Next**.
Choose your **Action Parameters**:
Mandatory arguments cannot be deselected.
1. Choose which arguments and inputs to include in the action. If a content item contains descriptions of the inputs, the descriptions are prepopulated. If not, you can provide short descriptions. The descriptions help the agent understand the purpose of each input.
2. (Optional) Enter a default value for each input. The default value is used when the user does not specify the input.
3. Choose which script outputs to include in the action. If the content item contains descriptions of the outputs, the descriptions are prepopulated. If not, you can provide short descriptions. The descriptions help the agent understand the purpose of each output.
**Save changes**.
The execution of system or custom actions that are based on integration commands can be restricted using [integration permissions](../../../reference-and-developer-docs/role-based-access-control/configuration-permissions/integrations-permissions).