pre-flight

Verify AI agent actions and plans against formalized policies before execution.

2|Updated Mar 24, 2026
One-click install
npx skills add https://github.com/ICME-Lab/icme-preflight-guardrail --skill pre-flight-icme-lab
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pre-flight
Source: https://github.com/ICME-Lab/icme-preflight-guardrail/tree/main
Command: npx skills add https://github.com/ICME-Lab/icme-preflight-guardrail --skill pre-flight-icme-lab

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill prevents AI agents from performing unapproved actions by verifying each step against formalized policies before execution.

Core Features & Use Cases

  • Logical Contradiction Detection: Quickly identifies conflicting reasoning or plans to prevent logical errors.
  • Policy Relevance Screening: Determines if an agent's proposed action relates to the defined policies, reducing unnecessary checks.
  • Action Verification: Checks specific actions against custom policies to enforce constraints on sensitive operations like data transfer, deletion, or external calls.
  • Use Case: You want to prevent an AI from exfiltrating data or executing forbidden commands—this Skill verifies each intent conforms to your rules and halts unsafe actions.

Quick Start

To verify whether a plan contains contradictions, call checkLogic with your reasoning; for each action, first use checkRelevance and then checkIt if relevant.

Frequently Asked Questions about pre-flight

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

FAQPage Schema
How do I prevent AI agents from executing unauthorized actions in automated workflows?

To prevent unauthorized actions, you can enforce formal verification policies that check each agent step against user-specified constraints before execution. This halts unsafe operations like data exfiltration by verifying intent conforms to your rules.

How do I check an AI agent's plan for logical contradictions?

To check an agent's plan for logical contradictions, call the logic verification function with your reasoning. This quickly identifies conflicting reasoning or plans to prevent logical errors before the agent proceeds with execution.

Can I screen AI agent actions for policy relevance before running full verification?

Yes, you can screen AI agent actions for policy relevance first. This determines if a proposed action relates to your defined policies, reducing unnecessary checks by filtering out irrelevant operations before full verification.

What is the best way to enforce constraints on sensitive AI operations like data transfer?

The best way to enforce constraints on sensitive AI operations is to verify specific actions against custom policies. By checking each intent against formalized rules, you can block forbidden commands, data deletion, or external calls.

Do I need formal logic to verify AI safety in automation workflows?

Yes, implementing formal logic ensures soundness when verifying AI safety in automation workflows. It provides the logical framework required to detect contradictions and enforce strict policy compliance before any agent action is executed.

When should I not use formal policy enforcement for AI agents?

Formal policy enforcement may not suit workflows requiring rapid, unblocked execution without compliance checks. If your automation operates in a fully trusted environment with no risk of unauthorized data transfer or forbidden commands, it is unnecessary.