nemo-guardrails

Configure guardrail middleware for LLM chat completions via NeMo Platform Inference Gateway.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill addresses the critical need for content safety and behavioral control in LLM-powered agents by preventing harmful, off-topic, or unauthorized responses.

Core Features & Use Cases

  • Content Safety: Implements input and output filtering against a defined safety taxonomy to block prohibited content.
  • VirtualModel Middleware: Integrates guardrail configurations directly into the inference pipeline via the NeMo Platform IGW.
  • Use Case: A customer support agent can use this Skill to ensure that the model never provides unauthorized financial advice or uses profanity, while also verifying that user inputs do not contain malicious jailbreak attempts.

Quick Start

Use the nemo-guardrails skill to create a new guardrail configuration named content-safety and attach it to your virtual model as request and response middleware.

Frequently Asked Questions about nemo-guardrails

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

FAQPage Schema
How do I add safety guardrails to LLM chat completions?

Configuring guardrail middleware enforces safety policies and behavioral constraints on LLM chat completions. This process filters inputs and outputs against a defined safety taxonomy to block prohibited content during inference.

What is the best way to prevent jailbreak attempts in LLM agents?

Preventing jailbreak attempts in LLM agents is best achieved by implementing input and output filtering against a defined safety taxonomy. This middleware blocks malicious user inputs and unauthorized model responses before they complete inference.

How do I configure middleware to block unauthorized responses in LLM inference?

Configuring middleware to block unauthorized responses involves attaching a guardrail configuration to your virtual model. This applies request and response middleware filters directly into the inference pipeline via the NeMo Platform Inference Gateway.

Does NeMo guardrail middleware support streaming response validation for LLMs?

Yes, guardrail middleware supports streaming response validation for LLMs. This allows the system to enforce safety policies and behavioral constraints dynamically during complex workflows and task-specific LLM calls.

Do I need the NeMo Platform Inference Gateway to use guardrails for LLM compliance?

Yes, integrating guardrails for LLM compliance requires the NeMo Platform Inference Gateway. The system uses the VirtualModel middleware registry to attach guardrail configurations directly into the inference pipeline.