What is a Rule of Constraints (RoC)?

A Rule of Constraints (RoC) is a policy bundle that defines the operational boundaries, allowed tools, and required approvals for an AI agent within a specific bounded context.

How does this skill improve AI safety?

It allows developers to explicitly forbid dangerous tools and mandate human review for specific triggers, preventing unintended actions by autonomous agents.

Where are the governance policies stored?

Policies are generated as YAML files within the runtime/contexts/ directory of your project, making them easy to version-control and audit.

Can I customize escalation triggers?

Yes, you can define specific conditions or 'keys' that trigger an escalation to a human or a specific supervisor role when the agent encounters a restricted scenario.

Does this skill support different agent roles?

Yes, you can specify agent types, models, and roles (like Strategist) to tailor the constraints to the specific function the AI is performing.

Agent Governance & Constraints

Name: Agent Governance & Constraints
Author: venikman

byvenikman

0•

生产力与工作流

Defines specialized Rule of Constraints (RoC) policy bundles to establish guardrails and autonomy boundaries for AI agents.

This skill automates the creation of Rule of Constraints (RoC) policy documents, allowing developers to precisely define the operational boundaries of AI agents within a specific context. By specifying allowed tools, forbidden actions, approval requirements, and escalation triggers, it ensures that autonomous agents operate safely, predictably, and within organizational compliance standards. It is an essential tool for teams looking to scale AI-driven workflows while maintaining strict oversight and security through structured policy enforcement.

主要功能

01Generate standardized YAML policy bundles for auditability and version control

02Configure mandatory approval triggers for sensitive or high-risk operations

03Create context-specific guardrails for bounded agent environments

040 GitHub stars

05Establish clear escalation paths for human intervention when constraints are met

06Define tool-use permissions including allowed and forbidden tool lists

使用场景

01Defining hierarchical approval workflows for automated infrastructure deployments

02Establishing safety boundaries for an AI agent performing sensitive database operations

03Restricting agent access to specific CLI tools in production-ready environments

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add venikman/fpf-foundry governance-define-roc

For use in Claude.ai and ChatGPT

Download Skill