writing-skills

Apply the RED-GREEN-REFACTOR cycle to create and refine AI agent skills.

18|8|Updated May 31, 2023
One-click install
npx skills add https://github.com/jpmorgan-payments/embedded-finance --skill writing-skills-jpmorgan-payments
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: writing-skills
Source: https://github.com/jpmorgan-payments/embedded-finance/tree/main/.github/skills/superpowers/writing-skills
Command: npx skills add https://github.com/jpmorgan-payments/embedded-finance --skill writing-skills-jpmorgan-payments

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill provides a rigorous, test-driven framework for creating and refining AI Skills, ensuring they are discoverable, effective, and resistant to rationalization, thereby improving AI agent performance and reliability.

Core Features & Use Cases

  • Skill Creation Framework: Guides users through the RED-GREEN-REFACTOR cycle for documentation.
  • Test Scenario Generation: Provides methods for creating pressure scenarios to test skill compliance.
  • Rationalization Countermeasures: Offers techniques to identify and block common excuses agents use to bypass rules.
  • Use Case: A team is developing a new Skill to enforce code review standards. They use this Skill to define test scenarios that simulate time pressure and authority, ensuring the new Skill is followed even when agents are tempted to skip steps.

Quick Start

Follow the RED-GREEN-REFACTOR process outlined in this skill to create a new skill.

Frequently Asked Questions about writing-skills

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

FAQPage Schema
How do I use TDD to create and refine AI agent skills?

You can use test-driven development for skill creation by applying the RED-GREEN-REFACTOR cycle to documentation. This enforces rigorous baseline testing and minimal skill writing to iteratively refine agent behavior.

What is the best way to prevent AI agents from rationalizing and bypassing rules?

The best way to prevent rationalization is to define pressure scenarios that test skill compliance under constraints like time pressure. This identifies and blocks common excuses agents use to bypass rules.

How does the RED-GREEN-REFACTOR cycle apply to documentation?

The RED-GREEN-REFACTOR cycle applies to documentation by requiring rigorous baseline tests first, writing minimal skill content to pass those tests, then iteratively refactoring to close loopholes and prevent rationalization.

Do I need any dependencies to apply test-driven development for skill creation?

No dependencies are required to apply this test-driven skill creation framework. It operates independently using built-in references to guide you through generating test scenarios and countermeasures.

Why should I test AI skills under pressure scenarios like time limits or authority?

Testing under pressure scenarios ensures skills remain effective and compliant even when agents are tempted to skip steps. It validates that your documentation successfully resists rationalization under stress.

What are the limitations of using a test-driven approach for skill documentation?

This approach requires rigorous baseline testing and iterative refactoring, which demands strict adherence to the cycle. You must continuously close loopholes in documentation to prevent agents from finding rationalization workarounds.