How does this prevent hallucinated success in AI coding?

It forces the AI to run real-time commands and read the full exit codes and output before it is allowed to use success-oriented language like 'Done', 'Fixed', or 'Passed'.

Can this skill be used for regression testing?

Yes, it includes specific patterns for TDD Red-Green cycles to ensure a fix is verified against the original symptom before being restored and re-verified.

When should I apply this skill in my workflow?

It should be applied before any commit, Pull Request creation, task delegation, or transition to a new task to ensure the current state is accurately reported.

What is the 'Iron Law' of this skill?

The Iron Law mandates that no completion claims can be made without fresh, documented verification evidence from running the actual commands required to prove the claim.

Verification Before Completion

Name: Verification Before Completion
Author: dkmaker

bydkmaker

•

安全与测试

Enforces a rigorous verification-first workflow that requires concrete evidence before any work is claimed as complete or successful.

This skill implements a strict 'Iron Law' for Claude, ensuring that it never reports a task as finished without first executing and validating fresh evidence through commands, tests, or builds. By mandating a systematic gate function—identify, run, read, and verify—it eliminates 'hallucinated success' and prevents shipping bugs, broken builds, or incomplete requirements. It is an essential tool for maintaining high trust and code quality by ensuring that no shortcuts are taken during the final stages of the development lifecycle.

主要功能

01Automated red-flag detection for speculative language like 'should' or 'probably'

02Comprehensive verification checklists for tests, builds, and bug fixes

031 GitHub stars

04Anti-rationalization framework to prevent shortcuts and partial checks

05Systematic gate function for validating status before commits or PRs

06Strict 'Evidence-First' protocol for all completion claims

使用场景

01Validating complex environment builds and linter status during final task stages

02Ensuring all unit and integration tests pass before opening a Pull Request

03Verifying bug fixes through a complete red-green cycle before reporting success

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dkmaker/my-claude-plugins verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill