What types of skills benefit most from subagent testing?

Skills that enforce discipline (like TDD), have high compliance costs, or contradict immediate goals (like speed over quality) benefit most from this testing method.

How do you 'refactor' a Claude Code skill?

Refactoring involves identifying new excuses the AI uses to bypass rules and updating the skill with explicit negations, rationalization tables, and red flags to close those loopholes.

Why use the RED phase for documentation testing?

The RED phase establishes a baseline by watching the agent fail without the skill, ensuring the skill is designed to solve actual observed failures rather than hypothetical ones.

What is a pressure scenario in AI skill testing?

A pressure scenario is a realistic simulation that combines multiple constraints like time limits, sunk costs, and authority pressure to see if an AI agent will violate its documented rules.

Skill Testing with Subagents

Name: Skill Testing with Subagents
Author: ed3dai

byed3dai

•

Security & Testing

Applies a Red-Green-Refactor cycle to ensure Claude Code skills remain compliant under pressure and resist rationalization.

This skill provides a rigorous framework for validating custom Claude Code instructions using Test-Driven Development (TDD) principles. By treating process documentation like code, it guides developers through creating 'pressure scenarios'—simulations involving tight deadlines, sunk costs, and conflicting priorities—to verify that an AI agent will actually follow instructions when it is most tempted to ignore them. It provides a structured workflow to identify AI rationalization patterns and systematically close loopholes, resulting in bulletproof, high-discipline skills.

Key Features

01Pressure scenario generation using multi-factor constraints

02Explicit loophole closure and negation templates

0345 GitHub stars

04Red-Green-Refactor workflow for process documentation

05Rationalization tracking to identify AI excuse patterns

06Meta-testing protocols for documentation clarity validation

Use Cases

01Hardening TDD enforcement rules to prevent AI from skipping tests

02Refining custom instructions to ensure durability in high-stakes environments

03Verifying complex compliance or security guidelines under pressure

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ed3dai/ed3d-plugins testing-skills-with-subagents

For use in Claude.ai and ChatGPT

Download Skill