gws-modelarmor-sanitize-response

googleworkspace/cli

Sanitize model responses through Google Model Armor templates for outbound safety.

What is gws-modelarmor-sanitize-response?

This skill applies Google Model Armor templates to sanitize model-generated responses before sending them to users. Use it when you need to filter or clean model outputs for safety compliance and content policy adherence.

Applies Model Armor templates to sanitize model responses
Accepts text input via --text flag or piped stdin
Supports full JSON request body override with --json flag
Processes outbound safety filtering (model to user)
Requires template resource name in projects/PROJECT/locations/LOCATION/templates/TEMPLATE format

How to install gws-modelarmor-sanitize-response

npx skills add null --skill gws-modelarmor-sanitize-response

Prerequisites

Google Workspace CLI (gws) installed and configured
Valid Google Cloud authentication via gws-shared setup
Access to a Model Armor template in your Google Cloud project
Template resource name: projects/PROJECT/locations/LOCATION/templates/TEMPLATE

Claude Code

Cursor

Windsurf

Cline

How to use gws-modelarmor-sanitize-response

1.Ensure you have a Model Armor template created in your Google Cloud project
2.Run the command with your template resource name: gws modelarmor +sanitize-response --template projects/PROJECT/locations/LOCATION/templates/TEMPLATE
3.Provide text to sanitize either via --text flag or pipe input from another command
4.Review the sanitized output returned by the template

Use cases

Good for

Filter model-generated content before displaying to end users
Ensure compliance with content policies on model outputs
Sanitize responses in multi-turn conversations before delivery
Remove sensitive or harmful content from AI-generated text
Apply organization-specific safety rules to model responses

Who it's for

AI application developers
Content safety engineers
Teams deploying LLM applications
Organizations with content moderation requirements

gws-modelarmor-sanitize-response FAQ

What's the difference between sanitize-response and sanitize-prompt?

sanitize-response filters model outputs (outbound safety, model to user), while sanitize-prompt filters user inputs (inbound safety, user to model).

Can I use this without a Model Armor template?

No, a valid template resource name is required. You must create a template in your Google Cloud project first.

How do I pass input to this skill?

Use the --text flag for direct text input, pipe content from another command, or use --json to provide a full JSON request body.

What format should the template resource name be?

Use the full format: projects/PROJECT/locations/LOCATION/templates/TEMPLATE where PROJECT, LOCATION, and TEMPLATE are your actual values.

Full instructions (SKILL.md)

Source of truth, from googleworkspace/cli.

name: gws-modelarmor-sanitize-response description: "Google Model Armor: Sanitize a model response through a Model Armor template." metadata: version: 0.22.5 openclaw: category: "security" requires: bins: - gws cliHelp: "gws modelarmor +sanitize-response --help"

modelarmor +sanitize-response

PREREQUISITE: Read ../gws-shared/SKILL.md for auth, global flags, and security rules. If missing, run gws generate-skills to create it.

Sanitize a model response through a Model Armor template

Usage

gws modelarmor +sanitize-response --template <NAME>

Flags

Flag	Required	Default	Description
`--template`	✓	—	Full template resource name (projects/PROJECT/locations/LOCATION/templates/TEMPLATE)
`--text`	—	—	Text content to sanitize
`--json`	—	—	Full JSON request body (overrides --text)

Examples

gws modelarmor +sanitize-response --template projects/P/locations/L/templates/T --text 'model output'
model_cmd | gws modelarmor +sanitize-response --template ...