When deploying Cortex XDR agent on Kubernetes clusters as a DaemonSet, the license required is Cortex XDR Cloud per Host. This license allows you to protect and monitor your cloud workloads, such as Kubernetes clusters, containers, and serverless functions, using Cortex XDR. With Cortex XDR Cloud per Host license, you can deploy Cortex XDR agents as DaemonSets on your Kubernetes clusters, which ensures that every node in the cluster runs a copy of the agent. The Cortex XDR agent collects and sends data from the Kubernetes cluster, such as pod events, container logs, and network traffic, to the Cortex Data Lake for analysis and correlation. Cortex XDR can then detect and respond to threats across your cloud environment, and provide visibility and context into your cloud workloads. The Cortex XDR Cloud per Host license is based on the number of hosts that run the Cortex XDR agent, regardless of the number of containers or functions on each host. A host is defined as a virtual machine, a physical server, or a Kubernetes node that runs the Cortex XDR agent. You can read more about the Cortex XDR Cloud per Host license and how to deploy Cortex XDR agent on Kubernetes clusters here1 and here2. Reference:
Cortex XDR Cloud per Host License
Deploy Cortex XDR Agent on Kubernetes Clusters as a DaemonSet

The difference between presets and datasets in XQL is that a dataset is a built-in or third-party data source, while a preset is a group of XDR data fields. A dataset is a collection of data that you can query and analyze using XQL. A dataset can be a Cortex data lake data source, such as endpoints, alerts, incidents, or network flows, or a third-party data source, such as AWS CloudTrail, Azure Activity Logs, or Google Cloud Audit Logs. A preset is a predefined set of XDR data fields that are relevant for a specific use case, such as process execution, file operations, or network activity. A preset can help you simplify and standardize your XQL queries by selecting the most important fields for your analysis. You can use presets with any Cortex data lake data source, but not with third-party data sources. Reference:
Datasets and Presets
XQL Language Reference

Phishing is a technique that belongs to two MITRE ATT&CK tactics: Reconnaissance and Initial Access. Reconnaissance is the process of gathering information about a target before launching an attack. Phishing for information is a sub-technique of Reconnaissance that involves sending phishing messages to elicit sensitive information that can be used during targeting. Initial Access is the process of gaining a foothold in a network or system. Phishing is a sub-technique of Initial Access that involves sending phishing messages to execute malicious code on victim systems. Phishing can be used for both Reconnaissance and Initial Access depending on the objective and content of the phishing message. Reference:
Phishing, Technique T1566 - Enterprise | MITRE ATT&CK 1
Phishing for Information, Technique T1598 - Enterprise | MITRE ATT&CK 2 Phishing for information, Part 2: Tactics and techniques 3 PHISHING AND THE MITREATT&CK FRAMEWORK - EnterpriseTalk 4 Initial Access, Tactic TA0001 - Enterprise | MITRE ATT&CK 5