Gws Modelarmor Create Template

What it does

This skill enables AI agents to create Model Armor templates within Google Workspace, specifically for protecting sensitive data used in generative AI applications. It allows users to define and deploy security configurations that act as filters for user prompts and responses. The skill creates these templates using the Google Cloud Platform (GCP), requiring a project ID and location.

When to use it

• When you need to establish standardized security protocols for generative AI applications within your organization.
• To create reusable templates for filtering potentially harmful or sensitive information in user prompts and responses.
• When deploying new generative AI applications that require pre-defined data protection measures.
• To quickly set up a baseline level of safety using the "jailbreak" preset template.

Key capabilities

• Creates Model Armor templates within GCP projects.
• Supports specifying project ID, location, and template ID as required parameters.
• Allows for the use of pre-defined presets like "jailbreak."
• Enables custom configuration through a JSON body (overriding presets).

Example prompts

• "Create a Model Armor template named 'my-tmpl' in project 'P' located in 'us-central1' using the jailbreak preset."
• "Generate a new Model Armor template with ID 'my-tmpl' in project 'P', region 'us-central1', and configure it using this JSON: '{...}'"

Tips & gotchas

• This is a "write" command, so always confirm the action with the user before execution.
• If you don’t specify --preset or --json, the skill defaults to the "jailbreak" preset.
• After creation, use the resulting template name with other Model Armor commands like +sanitize-prompt and +sanitize-response.