Can I use this for PII protection?

Yes, it includes specialized patterns for detecting and redacting sensitive information such as emails, phone numbers, and SSNs from LLM responses.

How does it help with LLM hallucinations?

It provides factuality-checking rails that verify LLM claims against source documents using Colang 2.0 flows and grounding validation techniques.

Is it compliant with OWASP security standards?

The skill is explicitly mapped to the OWASP Top 10 for LLM Applications 2025, providing mitigations for risks like excessive agency and insecure output handling.

Does it protect against prompt injection?

Yes, it implements specific input validation rails and red-teaming workflows to detect and mitigate prompt injection and jailbreaking attempts.

What frameworks does this skill support?

This skill integrates NeMo Guardrails, Guardrails AI, OpenAI moderation APIs, and DeepTeam for comprehensive LLM safety and adversarial testing.

Advanced LLM Guardrails & Safety

Name: Advanced LLM Guardrails & Safety
Author: yonatangross

byyonatangross

•

セキュリティとテスト

Implements production-grade LLM safety layers using NeMo Guardrails, Guardrails AI, and adversarial red-teaming to prevent hallucinations, toxicity, and prompt injections.

Advanced Guardrails provides a comprehensive security toolkit for developers building production-ready AI applications that require strict safety and reliability. By integrating industry-standard frameworks like NeMo Guardrails and Guardrails AI, this skill enables multi-layered input/output validation, PII redaction, and domain-specific restrictions. It is specifically designed to mitigate the OWASP Top 10 for LLMs, offering specialized patterns for hallucination prevention, fact-checking in RAG workflows, and automated red-teaming to identify vulnerabilities before deployment.

主な機能

01Compliance mapping for the OWASP Top 10 for LLM Applications

02Real-time hallucination detection and fact-checking for RAG systems

03Adversarial red-teaming and jailbreak detection using DeepTeam patterns

04Automated PII detection and redaction to prevent sensitive data disclosure

05Multi-layer input and output validation using NeMo Guardrails and Guardrails AI

0669 GitHub stars

ユースケース

01Securing enterprise chatbots against toxic language, off-topic queries, and prompt injections

02Enforcing factuality and grounding for LLMs in high-stakes domains like finance or healthcare

03Automating security audits and vulnerability scanning for AI agents prior to production release

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit advanced-guardrails

For use in Claude.ai and ChatGPT

Download Skill