Gws Modelarmor

What it does

Gws Modelarmor enhances security and compliance within Google Workspace by automatically detecting and blocking risky user actions. It provides commands to sanitize both user prompts and model responses using predefined templates. The skill allows users to create new Model Armor templates for customized protection, ensuring data safety when interacting with AI models.

When to use it

• When you need to ensure user prompts sent to an AI model are safe and compliant.
• To sanitize the responses generated by a model before they are presented to a user.
• When creating custom security rules or templates for protecting Google Workspace data.
• For initial setup of Model Armor configurations within your environment.

Key capabilities

• sanitize-prompt: Sanitizes user prompts using a Model Armor template.
• sanitize-response: Sanitizes model responses using a Model Armor template.
• create-template: Creates new Model Armor templates.
• Provides schema discovery to understand API method parameters and types.

Example prompts

• "gws modelarmor users.list --params='{"query": "[email protected]"}'" (This is an example of how to use the skill, not a direct prompt for ModelArmor itself)
• "gws modelarmor +sanitize-prompt 'User input here'"
• "gws modelarmor +create-template --name 'My New Template' --description 'A custom template.'"

Tips & gotchas

• Prerequisite: You must have a SKILL.md file for authentication, global flags, and security rules. Use gws generate-skills to create this if it’s missing.
• The skill uses command-line arguments with --params and --json flags which need to be built using the schema output (gws schema modelarmor.<resource>.<method>).
• Before using any API method, use gws modelarmor --help to browse resources and methods.