custodiet

Detect AI agent actions exceeding granted authority by analyzing input files.

3|4|Updated Mar 11, 2026

Search Tags:#compliance #policy enforcement #guardrails #ai governance #agent security #ultra vires

Ready to run

npx skills add https://github.com/nicsuzor/academicOps --skill custodiet

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: custodiet
Source: https://github.com/nicsuzor/academicOps/tree/main/.agent/skills/custodiet
Command: npx skills add https://github.com/nicsuzor/academicOps --skill custodiet

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill prevents AI agents from performing actions that exceed their authorized operational boundaries, ensuring they adhere to defined rules and user intent.

Core Features & Use Cases

Ultra Vires Detection: Identifies and flags actions taken by an AI agent that go beyond its granted permissions or scope.
Compliance Enforcement: Ensures AI operations align with established framework principles (axioms, heuristics).
Use Case: An agent is tasked with summarizing a document. If it attempts to modify system files or access restricted data, Custodiet will detect this "ultra vires" action and block it.

Quick Start

Use the custodiet skill to check if the agent's actions in the provided file path are within its granted authority.

Frequently Asked Questions about custodiet

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

FAQPage Schema

How do I detect when an AI agent operates beyond its granted authority?▼

To detect ultra vires actions, you analyze the agent's input files against established framework principles. This compliance monitoring process identifies and flags actions that exceed the agent's defined operational boundaries and permissions.

What is ultra vires detection in AI agent governance?▼

Ultra vires detection in AI governance identifies and flags actions taken by an autonomous agent that go beyond its granted permissions or scope. It ensures agents adhere to defined rules and user intent within their operational zones.

How do I enforce compliance and guardrails for autonomous AI agents?▼

You enforce compliance and guardrails for autonomous AI agents by applying framework principles like axioms and heuristics to monitor their operations. This ensures alignment with established rules and prevents unauthorized actions within defined operational zones.

Do I need specific output formats for programmatic AI policy enforcement?▼

Yes, programmatic AI policy enforcement requires adherence to specific output formats for parsing and enforcement actions. This structured output allows downstream systems to automatically process detected ultra vires actions and block them programmatically.

Can I use a file path to check if agent actions are within granted authority?▼

Yes, you can check if agent actions are within granted authority by providing a file path containing the agent's operational logs. The system analyzes these input files against framework principles to detect any ultra vires actions.

What are the limitations of using framework principles for agent security?▼

The limitation of using framework principles for agent security is that detection depends entirely on the defined axioms and heuristics provided. Actions not explicitly covered by these established rules may not be flagged as ultra vires.