Can this skill help with PII protection?

Yes, it includes specialized patterns for detecting and redacting Personally Identifiable Information (PII) such as emails, phone numbers, and SSNs in both inputs and outputs.

What frameworks does this skill support?

This skill supports NeMo Guardrails (for programmable flows), Guardrails AI (for validator-based checks), OpenAI Moderation, and DeepTeam for adversarial testing.

How does it prevent hallucinations?

It implements fact-checking rails and grounding checks using Colang 2.0 and NLI-based validators to verify LLM claims against source documentation.

Is it compatible with OWASP security standards?

Yes, it provides a direct mapping to the OWASP Top 10 for LLMs 2025, offering specific guardrail solutions for risks like prompt injection and sensitive info disclosure.

Advanced LLM Guardrails & Safety

Name: Advanced LLM Guardrails & Safety
Author: yonatangross

byyonatangross

•

Security & Testing

Implements comprehensive safety layers for LLM applications using NeMo, Guardrails AI, and red-teaming validation patterns.

Advanced Guardrails is a production-ready toolkit designed to secure LLM applications by enforcing strict input and output validation. It integrates industry-standard frameworks like NeMo Guardrails and Guardrails AI to prevent hallucinations, redact sensitive PII, and block toxic content. By providing pre-configured Colang 2.0 flows and OWASP Top 10 compliance patterns, it enables developers to build trustworthy AI systems that adhere to domain-specific topics while defending against adversarial attacks and prompt injections.

Key Features

01Automated PII detection and redaction for sensitive data protection

02Comprehensive red-teaming and adversarial testing modules

03Multi-framework support for NeMo Guardrails, Guardrails AI, and OpenAI

04Hallucination prevention and fact-checking for RAG-based systems

0569 GitHub stars

06OWASP Top 10 for LLMs security compliance mapping

Use Cases

01Auditing LLM implementations for security vulnerabilities and prompt injection risks.

02Ensuring factual accuracy in RAG applications for specialized domains like health or finance.

03Building enterprise-grade AI chatbots that require strict topic adherence and PII protection.

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