exo:guardrails

Detect and block prompt injections for Exo agents with guardrails.

1|Updated Feb 16, 2026
One-click install
npx skills add https://github.com/midsphere-ai/exo --skill exo-guardrails
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: exo:guardrails
Source: https://github.com/midsphere-ai/exo/tree/main/skills/guardrails
Command: npx skills add https://github.com/midsphere-ai/exo --skill exo-guardrails

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Security guardrails for Exo agents to detect and block prompt injection, enforce content filtering, and govern tool usage and workflow decisions.

Core Features & Use Cases

  • Pattern-based input protection (fast checks)
  • LLM-backed risk analysis for deep detection
  • Rails-based control flow (CONTINUE/SKIP/RETRY/ABORT) across tool calls and LLM interactions
  • Safe fail-open behavior and cross-rail coordination for robust safety
  • Real-world scenarios: preventing leakage of system prompts, blocking dangerous tool calls, and enforcing policy checks

Quick Start

Attach a guardrail to the agent to block common prompt-injection patterns before the LLM call.

Frequently Asked Questions about exo:guardrails

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

FAQPage Schema
How do I prevent prompt injection attacks on LLM agents?

Prevent prompt injection on LLM agents by applying pattern-based input protection and LLM-backed risk analysis to detect and block malicious inputs before the LLM call. This approach secures agent deployments against system prompt leakage.

How do I enforce guardrails to block dangerous tool calls in production?

Enforce guardrails for dangerous tool calls by applying Rails-based control flow to govern decisions across tool invocations and LLM interactions. This allows you to define explicit CONTINUE, SKIP, RETRY, or ABORT behaviors for safe agent operation.

What is the best way to detect prompt injections in real-time agent workflows?

The best way to detect prompt injections in real-time workflows is combining fast pattern-based input checks with deep LLM-backed risk analysis. This dual-layer detection validates inputs and moderates content across pre-LLM calls and decision-flow control.

Can I use LLM guardrails to enforce content filtering and policy checks?

Yes, you can use LLM guardrails to enforce content filtering and policy checks by integrating BaseGuardrail and Rail systems. These systems validate inputs and govern tool usage to ensure agent compliance with defined safety policies.

What happens when an LLM guardrail fails during content moderation?

When an LLM guardrail fails during content moderation, the system executes safe fail-open behavior and cross-rail coordination. This ensures robust safety by defining explicit fallback behaviors to handle failures without compromising the workflow.