Why can't Claude just say the code 'looks correct'?

This skill follows the principle that 'evidence comes before assertions.' Visual inspection is insufficient; only actual execution logs constitute valid proof of completion.

How does this prevent trust issues in development?

By providing real output and exit codes for every claim, it removes the 'I don't believe you' friction between the AI and the human developer.

How does it handle bug fixes?

It mandates a red-green cycle: the agent must prove the test fails without the fix and passes with it to confirm the regression is truly resolved.

What is the primary purpose of this skill?

The skill ensures that Claude never claims a task is done without first running and confirming the output of a relevant verification command, such as a test or build.

Does this skill work with any programming language?

Yes, it is language-agnostic as it relies on the execution of shell commands (npm, pytest, cargo, etc.) to verify the state of the project.

Verification Before Completion

Name: Verification Before Completion
Author: ed3dai

byed3dai

•

セキュリティとテスト

Enforces rigorous evidence-based reporting by requiring fresh verification command output before any task is marked as finished.

The Verification Before Completion skill establishes a mandatory 'Gate Function' for Claude to ensure code quality and integrity. It prevents premature or dishonest success claims by requiring the AI to identify, execute, and analyze specific verification commands—such as test suites, linters, and build scripts—before asserting that a task is complete. By prioritizing fresh evidence over assumptions or confidence, this skill eliminates common failures like shipping undefined functions or broken regressions, ensuring every claim is backed by reproducible logs and exit codes.

主な機能

01Comprehensive validation checklist for tests, builds, and linting

02Iron Law enforcement to prevent 'should pass' rationalizations

03Red-green regression cycle verification for robust bug fixes

045 GitHub stars

05Independent verification of changes made by delegated agents

06Mandatory execution of verification commands before success assertions

ユースケース

01Verifying bug fixes through documented failing and passing test cycles

02Validating build stability and requirement fulfillment in complex workflows

03Ensuring 100% test pass rates before creating a Pull Request

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ed3dai/ed3d-plugins-testing verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill