When should I avoid creating a skill?

Do not create skills for one-off solutions, project-specific conventions, or standard practices that are already well-documented elsewhere.

Why do I need a failing test first?

Without a baseline failure, you cannot verify if the skill actually improves agent performance or if the agent is simply rationalizing its behavior.

How do I test a skill with subagents?

Create pressure scenarios where the agent is likely to fail, document the baseline error, then introduce the skill to verify it consistently fixes the issue.

What is TDD for skills?

It is the application of Test-Driven Development principles to documentation, requiring you to witness an agent failure before writing the instruction to fix it.

Skill Authoring TDD

Name: Skill Authoring TDD
Author: NickCrew

byNickCrew

•

학습 및 문서화

Standardizes the creation of robust AI skills through a Test-Driven Development approach to ensure agent compliance.

소개

The Writing Skills plugin implements a rigorous 'RED-GREEN-REFACTOR' workflow specifically designed for creating AI process documentation and agent instructions. By requiring developers to witness agent failure in pressure scenarios before writing any documentation, it ensures that every skill addresses a specific, observed failure mode. This methodology eliminates agent rationalization, closes logical loopholes, and results in bulletproof, high-performance skills for Claude Code and other agentic systems.

주요 기능

Structured organization for technique and pattern skills
Mandatory verification and deployment checklists
Anti-rationalization loophole closing
7 GitHub stars
RED-GREEN-REFACTOR documentation workflow
Baseline behavior testing using subagents

사용 사례

Creating new domain-specific AI skills for Claude Code
Fixing inconsistent agent behavior in complex technical workflows
Verifying documentation efficacy through pressure testing before deployment

소개

주요 기능

Structured organization for technique and pattern skills
Mandatory verification and deployment checklists
Anti-rationalization loophole closing
7 GitHub stars
RED-GREEN-REFACTOR documentation workflow
Baseline behavior testing using subagents

사용 사례

Creating new domain-specific AI skills for Claude Code
Fixing inconsistent agent behavior in complex technical workflows
Verifying documentation efficacy through pressure testing before deployment