gws-modelarmor-create-template

What This Skill Does

The gws-modelarmor-create-template skill allows developers and administrators to programmatically define safety templates within Google Cloud's Model Armor service. Model Armor acts as a defensive layer for AI applications, filtering prompts and responses for harmful content, jailbreak attempts, and policy violations. By utilizing this skill, users can standardize their safety configurations, ensuring that all AI interactions across their project adhere to consistent security guardrails. This skill automates the instantiation of these templates in specific GCP regions, facilitating infrastructure-as-code workflows for LLM safety management.

Installation

To integrate this skill into your environment, run the following installation command in your terminal:

clawhub install openclaw/skills/skills/googleworkspace-bot/gws-modelarmor-create-template

Ensure that you have completed the prerequisite setup defined in gws-shared/SKILL.md regarding authentication, as this skill interacts directly with Google Cloud Platform APIs and requires valid project permissions.

Use Cases

This skill is ideal for teams deploying production-grade generative AI models. Use it to:

• Deploy standardized 'jailbreak' protection presets across development, staging, and production environments.
• Create custom safety configuration templates that reflect organizational compliance policies.
• Programmatically update security postures without relying on the GCP web console.
• Automate safety auditing by applying template IDs across various prompt and response sanitization endpoints.

Example Prompts

1. "Create a new Model Armor template for the production project 'ai-safety-prod' in 'us-central1' with the ID 'standard-guardrails' using the jailbreak preset."
2. "Set up a custom Model Armor template named 'finance-bot-security' in 'europe-west1' using the provided JSON configuration file."
3. "Can you help me initialize a new Model Armor template with the ID 'testing-temp' in the 'us-east1' region for my current project?"

Tips & Limitations

• Write Operation: As this is a write-heavy operation that alters your cloud infrastructure, the system will prompt for confirmation. Always double-check your project ID and location before executing.
• Defaults: If you do not specify a --preset or --json configuration, the tool defaults to a standard jailbreak protection template.
• Hierarchy: Refer to the gws-modelarmor documentation to understand how these templates interact with the +sanitize-prompt and +sanitize-response commands.
• Validation: Ensure your JSON schema matches the requirements for Model Armor templates, as invalid JSON will cause the API call to fail during execution.