How does this prevent AI hallucinations?

It forces the AI to execute terminal commands and inspect raw exit codes and output before it is allowed to state that a task is finished, preventing it from assuming code works without running it.

When should this skill be applied?

It should be applied before claiming success, expressing satisfaction with a result, moving to a new task, or committing/pushing code to a repository.

Can I use this for regression testing?

Yes, the skill specifically mandates a Red-to-Green cycle—demonstrating a failure, applying a fix, and then showing a pass—to prove a bug is actually resolved.

What is the 'Iron Law' of this skill?

The Iron Law states that no completion claims are permitted without fresh, verifiable evidence from a command run performed in the current message context.

Engineering Verification

Name: Engineering Verification
Author: tjboudreaux

bytjboudreaux

•

安全与测试

Enforces an evidence-based gate function that requires fresh command output and validation before any task is marked as complete.

The Engineering Verification skill introduces a rigorous 'Iron Law' for software development within Claude Code: no completion claims without fresh verification evidence. It provides a systematic gate function that prevents 'hallucinated' success by requiring the AI to run specific commands, inspect exit codes, and present raw output before reporting a task as finished. This skill is essential for maintaining high-quality standards in TDD, bug fixing, and automated refactoring by eliminating reliance on assumptions, 'should work' statements, or outdated test runs.

主要功能

01Evidence-based reporting requirements for tests, linting, and builds

02Structured validation checklist for agent-driven code changes

03Mandatory verification gate function for all completion claims

04Red-to-Green cycle verification for regression testing

051 GitHub stars

06Anti-pattern detection for vague or unsubstantiated status updates

使用场景

01Ensuring unit tests actually pass before submitting a pull request

02Validating build and lint status after complex refactoring tasks

03Verifying that a bug fix includes a reproducible failing test that now passes

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tjboudreaux/cc-plugin-engineering-excellence eng-verification

For use in Claude.ai and ChatGPT

Download Skill