Does this guide help with token costs?

Absolutely. Principle 1 (Context Efficiency) focuses on providing maximum information with minimum tokens, directly reducing the cost of long-running AI sessions.

Can I use these rules for existing agent prompts?

Yes, the guide is specifically designed for creating or modifying skills, agent prompts, and command instructions to ensure consistent execution.

Why should I include NG patterns in my rules?

NG (No-Good) patterns provide clear examples of what the AI should avoid, which is often more memorable and effective for LLMs than abstract instructions alone.

What is the Rule Editing Guide?

It is a set of nine standardized principles designed to help developers write more effective, token-efficient, and accurate instructions for AI agents within Claude Code.

How does this skill improve AI performance?

By using techniques like unified notation, measurable criteria, and NG patterns, it reduces ambiguity and ensures the LLM follows complex instructions more reliably.

Rule Editing Guide

Name: Rule Editing Guide
Author: tundraray

bytundraray

•

生産性とワークフロー

Optimizes LLM execution accuracy through nine core principles for writing and refining Claude Code skills and agent instructions.

The Rule Editing Guide provides a structured framework for developers to create high-performance instructions for AI agents. By implementing nine foundational principles—including context efficiency, measurable criteria, and unified notation—this skill ensures that custom rules, agent prompts, and command definitions are interpreted with maximum precision. It helps eliminate ambiguity, reduce token consumption, and create a consistent logic layer for Claude Code workflows, making it essential for teams scaling their AI-driven development processes.

主な機能

01Improves token efficiency through concise, high-impact rule writing

02Establishes clear scope boundaries and measurable success criteria for agent tasks

031 GitHub stars

04Standardizes instruction notation using MUST/SHOULD/MAY terminology

05Optimizes prompt structure by prioritizing critical constraints first

06Provides NG (No-Good) pattern comparisons to prevent common AI failures

ユースケース

01Standardizing team-wide coding principles and LLM interaction guidelines

02Refactoring existing agent prompts to reduce token usage and improve reliability

03Creating a new specialized skill for a specific development framework

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tundraray/overture rule-editing-guide

For use in Claude.ai and ChatGPT

Download Skill