What is the 'Knowledge Test' in skill design?

It is a evaluation criteria that asks 'Does this TEACH Claude or DO something?' to ensure the component is categorized correctly for optimal performance.

What is the main difference between a skill and an agent?

A skill teaches Claude patterns, conventions, and methodologies (TEACH), while an agent executes multi-step processes or produces artifacts autonomously (DO).

Why should I avoid persona statements in skills?

Persona statements like 'You are an expert' are considered decorative. Skills are more effective when they provide concrete patterns, decision criteria, and error handling guidance.

When is a skill considered 'low quality'?

A skill is low quality if it has a vague scope, is not actionable, includes ASCII art, or attempts to perform autonomous work that should be handled by an agent.

Skill Design Philosophy

Name: Skill Design Philosophy
Author: majesticlabs-dev

bymajesticlabs-dev

•

生産性とワークフロー

Defines core principles for building effective Claude Code skills by distinguishing between teaching patterns and executing tasks.

概要

This skill serves as a definitive guide for developers building on the Claude Code platform, helping them determine when a component should be structured as a skill versus an autonomous agent. It introduces the 'Knowledge Test' to ensure skills focus on teaching Claude specific coding conventions, framework patterns, and methodologies rather than performing autonomous work. By applying these design principles, developers can create high-quality, actionable skills that improve Claude's behavior through context and specific guidance while eliminating decorative or redundant instructions.

主な機能

Provides concrete examples of high-quality vs. low-quality skill definitions
Applies the 'Knowledge Test' to evaluate component purpose
Distinguishes between TEACH (Skill) and DO (Agent) architectures
18 GitHub stars
Offers framework-specific idiom and pattern guidance
Defines content rules for eliminating decorative or redundant text

ユースケース

Deciding whether to build a new capability as a skill or an agent
Auditing existing Claude skills for quality and actionable content
Establishing standardized coding conventions for a development team

概要

主な機能

Provides concrete examples of high-quality vs. low-quality skill definitions
Applies the 'Knowledge Test' to evaluate component purpose
Distinguishes between TEACH (Skill) and DO (Agent) architectures
18 GitHub stars
Offers framework-specific idiom and pattern guidance
Defines content rules for eliminating decorative or redundant text

ユースケース

Deciding whether to build a new capability as a skill or an agent
Auditing existing Claude skills for quality and actionable content
Establishing standardized coding conventions for a development team