agent-governance

Implement governance controls for AI agents with declarative policies and audit trails.

1|Updated Feb 24, 2026
One-click install
npx skills add https://github.com/xurenlu/marstaff --skill agent-governance-xurenlu
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-governance
Source: https://github.com/xurenlu/marstaff/tree/main/skills/agent-governance
Command: npx skills add https://github.com/xurenlu/marstaff --skill agent-governance-xurenlu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill provides patterns and techniques to implement governance, safety, and trust controls for AI agent systems, ensuring they operate within defined boundaries and maintain accountability.

Core Features & Use Cases

  • Policy Enforcement: Define and apply granular policies for tool usage, content filtering, and rate limiting.
  • Threat Detection: Implement semantic intent classification to identify and block dangerous prompts.
  • Trust Scoring: Track agent reliability over time for multi-agent interactions.
  • Audit Trails: Maintain an immutable log of all agent actions for compliance and debugging.
  • Use Case: Securely enabling an AI agent to access a company database by defining a policy that only allows read operations on specific tables and blocks any commands containing sensitive keywords.

Quick Start

Use the agent-governance skill to define a policy that allows only the 'search' and 'summarize' tools and blocks any content containing the word 'password'.

Frequently Asked Questions about agent-governance

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

FAQPage Schema
How do I enforce policy controls for AI agents calling external tools?

To enforce policy controls for AI agents, you apply granular declarative policies and tool-level decorators that restrict operations, filter content, and enforce rate limits on external tool calls.

How does semantic intent classification work for agent threat detection?

Semantic intent classification detects threats by analyzing prompt meaning to identify and block dangerous inputs before AI agents execute harmful actions or unauthorized external tool calls.

Can I use declarative policies with CrewAI and PydanticAI frameworks?

Yes, declarative policies integrate with agent frameworks like CrewAI, PydanticAI, and OpenAI Agents SDK to govern multi-agent workflows and secure production deployments.

What is the best way to maintain an audit trail for AI agent compliance?

The best way to maintain an audit trail for AI agent compliance is to generate immutable logs of all agent actions, enabling accountability, debugging, and regulatory compliance tracking.

How do I calculate trust scoring for multi-agent interactions?

Trust scoring for multi-agent interactions tracks agent reliability over time, evaluating behavioral patterns to establish accountability and safety boundaries within production deployments.