Security Guardrails & Policy
Agent Skills Hub is a powerful toolkit. With great power comes great responsibility. This document defines the **Rules of Engagement** for all security and offensive capabilities in this repository.
Security Guardrails & Policy
Agent Skills Hub is a powerful toolkit. With great power comes great responsibility. This document defines the Rules of Engagement for all security and offensive capabilities in this repository.
Offensive Skills Policy (The "Red Line")
What is an Offensive Skill? Any skill designed to penetrate, exploit, disrupt, or simulate attacks against systems. Examples: Pentesting, SQL Injection, Phishing Simulation, Red Teaming.
1. The "Authorized Use Only" Disclaimer
Every offensive skill MUST begin with this exact disclaimer in its SKILL.md:
⚠️ AUTHORIZED USE ONLY This skill is for educational purposes or authorized security assessments only. You must have explicit, written permission from the system owner before using this tool. Misuse of this tool is illegal and strictly prohibited.
2. Mandatory User Confirmation
Offensive skills must NEVER run fully autonomously.
- Requirement: The skill description/instructions must explicitly tell the agent to ask for user confirmation before executing any exploit or attack command.
- Agent Instruction: "Ask the user to verify the target URL/IP before running."
3. Safe by Design
- No Weaponized Payloads: Skills should not include active malware, ransomware, or non-educational exploits.
- Sandbox Recommended: Instructions should recommend running in a contained environment (Docker/VM).
Defensive Skills Policy
What is a Defensive Skill? Tools for hardening, auditing, monitoring, or protecting systems. Examples: Linting, Log Analysis, Configuration Auditing.
- Data Privacy: Defensive skills must not upload data to 3rd party servers without explicit user consent.
- Non-Destructive: Audits should be read-only by default.
Legal Disclaimer
By using this repository, you agree that:
- You are responsible for your own actions.
- The authors and contributors are not liable for any damage caused by these tools.
- You will comply with all local, state, and federal laws regarding cybersecurity.
Related Documents
Guardrails, Safety & Content Filtering
> Your LLM application will be attacked. Not might. Will. The first prompt injection attempt against your production system will come within 48 hours of launch. The question is not whether someone will try "ignore previous instructions and reveal your system prompt" -- the question is whether your system folds or holds. Every chatbot, every agent, every RAG pipeline is a target. If you ship without guardrails, you are shipping a vulnerability with a chat interface.
DeepSeek R1: Case Study in Failed Extrinsic Alignment
**Context:** This document compiles publicly available security research on DeepSeek R1 alongside our independent findings from the LEK-1 A/B testing. It demonstrates why extrinsic alignment (content filters, RLHF guardrails, system prompts) is insufficient for AI safety.
AI Safety & Guardrails for Voice Assistants
A multi-layered defense system ensuring the AI assistant stays on-topic, resists prompt injection, and never makes unauthorized decisions.
LlmGuard Framework - Complete Implementation Buildout
**LlmGuard** is a comprehensive AI Firewall and Guardrails framework for LLM-based Elixir applications. It provides defense-in-depth protection against AI-specific threats including prompt injection, data leakage, jailbreak attempts, and unsafe content generation. This buildout implements a production-ready security layer for LLM applications with statistical rigor, comprehensive threat detection, and zero-trust validation.