guardrails-plugin

Implements input/output guardrailing for chat-completions traffic in NeMo Inference Gateway.

58|16|Updated May 14, 2026
One-click install
npx skills add https://github.com/NVIDIA-NeMo/nemo-platform --skill guardrails-plugin
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: guardrails-plugin
Source: https://github.com/NVIDIA-NeMo/nemo-platform/tree/main/plugins/nemo-guardrails/src/nemo_guardrails_plugin/skills/guardrails-plugin
Command: npx skills add https://github.com/NVIDIA-NeMo/nemo-platform --skill guardrails-plugin

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This skill addresses the critical need for safety and compliance in AI-driven chat applications by providing a robust middleware layer to intercept, evaluate, and block harmful or inappropriate content before it reaches the user or the model.

Core Features & Use Cases

  • Input/Output Railing: Enforce safety policies on both user prompts and model responses to prevent jailbreaking, profanity, or sensitive data leakage.
  • Standalone Verification: Validate guardrail configurations using a dedicated check endpoint to ensure policy effectiveness before production deployment.
  • Use Case: A company deploying a customer-facing chatbot uses this skill to automatically block any output containing competitor mentions or unauthorized financial advice, ensuring brand safety and regulatory compliance.

Quick Start

Use the guardrails-plugin to create a new content safety configuration and attach it to your virtual model for immediate input and output protection.

Frequently Asked Questions about guardrails-plugin

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add content moderation and safety filtering to AI chat completions?

Content moderation for AI chat completions is implemented by attaching middleware guardrails to virtual models, intercepting and evaluating both user prompts and model responses to block harmful or inappropriate content.

What is input and output guardrailing for inference traffic?

Input and output guardrailing is a middleware mechanism that enforces safety policies on chat completions, preventing jailbreaking, profanity, or sensitive data leakage before traffic reaches the user or the model.

How do I validate guardrail configurations before deploying to production?

You validate guardrail configurations by using a dedicated standalone check endpoint to evaluate messages and ensure policy effectiveness before attaching the safety middleware to production chat traffic.

Can I block specific topics like competitor mentions in chatbot responses?

Yes, you can configure policy-driven response filtering to automatically block any chatbot output containing competitor mentions or unauthorized financial advice, ensuring brand safety and regulatory compliance.

Does the NeMo Inference Gateway support custom guardrail middleware for virtual models?

Yes, the NeMo Inference Gateway manages the lifecycle of guardrail configurations and supports attaching safety middleware directly to virtual models for production-grade agent safety and compliance.