How does this skill prevent AI agents from 'cheating' on tests?

The skill enforces an Anti-Cheat Contract requiring the agent to provide exact test output snippets. It prohibits moving to the next phase until 'pass' or 'fail' evidence is physically shown in the chat.

What is 'Dependency Promotion' in this workflow?

If passing a test requires changing a separate module or dependency, the agent must stop and promote that dependency to the primary 'subject under test,' writing a new failing test for it before proceeding.

What happens if the agent cannot run the tests directly?

The skill enters a 'BLOCKED' state. The agent must explicitly state why it cannot proceed and request the user to run a specific command and provide the output manually.

Is this skill suitable for frontend and backend development?

Yes. While it focuses on logic and interfaces, the 'Outside-In' principle applies to REST endpoints, command handlers, and UI components alike, ensuring every piece of code has a testable entry point.

Strict TDD Workflow

Name: Strict TDD Workflow
Author: humansintheloop-dev

byhumansintheloop-dev

•

セキュリティとテスト

Enforces a rigorous, state-based Test-Driven Development (TDD) process to ensure AI-generated code is honest, minimal, and fully verified.

The TDD skill provides a disciplined framework for AI coding agents to follow the Red-Green-Refactor cycle with absolute integrity. It features a strict anti-cheat contract that prevents agents from inferring or hallucinating test results, requiring raw output evidence before any state transition. By promoting outside-in development and minimal implementation patterns, this skill ensures that production code is only written to satisfy specific failing tests, making it an essential tool for high-stakes software engineering where code reliability and architectural cleanliness are paramount.

主な機能

019 GitHub stars

02Outside-In Development: Prioritizes external interfaces to ensure all code is immediately reachable and testable.

03Dependency Promotion Rule: Ensures dependencies are tested as first-class units before integration.

04Minimal Implementation: Enforces the 'just enough code' rule to prevent feature creep and over-engineering.

05Strict Anti-Cheat Protocol: Mandates raw test output evidence for every state transition.

06State-Based Workflow: Guides agents through Planning, Red, Green, Refactor, and Verify phases.

ユースケース

01Onboarding AI agents to high-compliance projects that require verifiable proof of test execution.

02Developing complex backend logic where every edge case must be validated through unit tests.

03Refactoring legacy codebases by establishing a strict test-first safety net.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add humansintheloop-dev/humansintheloop-dev-workflow-and-tools tdd

For use in Claude.ai and ChatGPT

Download Skill