refine-boundaries

Add and evolve Always, Ask, and Never rules in AGENTS.md or CLAUDE.md.

1|Updated Feb 12, 2026
One-click install
npx skills add https://github.com/dev-goraebap/grimoire --skill refine-boundaries
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: refine-boundaries
Source: https://github.com/dev-goraebap/grimoire/tree/main/agent-cowork/skills/refine-boundaries
Command: npx skills add https://github.com/dev-goraebap/grimoire --skill refine-boundaries

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps teams progressively strengthen project-wide policies by accumulating Boundaries guidelines in the public instruction files (AGENTS.md or CLAUDE.md) so that future agents have a growing, additive rule set to reference instead of rewriting existing rules.

Core Features & Use Cases

  • Incremental rule augmentation: add a single rule per invocation under the appropriate section.
  • Flexible placement: inline within topic sections or as a cross-cutting Boundaries entry.
  • Auto source detection: prefer AGENTS.md, then CLAUDE.md, with optional track override.

Quick Start

Add a Never do rule to prevent a high-risk action.

Frequently Asked Questions about refine-boundaries

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

FAQPage Schema
How do I add project-wide policy rules for AI agents in AGENTS.md?

Yes, you can enforce testing and git workflow guidelines by adding Boundaries rules under specific topic sections. The skill accepts inline placement within topics like Code Style or Git Workflow, categorizing behaviors as Always, Ask, or Never for agents to follow.

Does this rule management approach preserve existing project rules when adding new ones?

This rule management approach preserves existing project rules by strictly applying additive-only updates. It never rewrites or deletes current boundaries, ensuring your accumulated policy set grows safely over time while maintaining all previously established guidelines.

Can I use CLAUDE.md instead of AGENTS.md for accumulating boundary guidelines?

You can use CLAUDE.md for accumulating boundary guidelines if AGENTS.md is absent. The skill features auto source detection that prefers AGENTS.md first, then falls back to CLAUDE.md, with an optional track override for custom file targeting.

What is the best way to prevent high-risk actions for AI agents in software development?

The best way to prevent high-risk actions is to define Never do rules within your project instruction files. This mechanism establishes explicit boundaries that restrict agents from executing dangerous operations across code style, testing, and git workflows.

How do boundaries and rule management work for AI agent behavior control?

Boundaries and rule management work by progressively accumulating policy guidelines in instruction files so future agents reference a growing, additive rule set. Instead of rewriting rules, it appends new behavioral constraints per invocation under relevant sections.