Does it work with automated agents?

Yes, it specifically mandates that reports from other agents must be independently verified by checking VCS diffs and running local tests rather than trusting the report blindly.

What kind of 'red flags' does it look for?

It flags language indicating uncertainty, such as 'should work,' 'probably,' or 'seems to,' as well as premature expressions of satisfaction before verification is run.

What is the primary purpose of this skill?

The skill ensures that the AI never claims a task is complete or fixed without first running a verification command and confirming the actual output.

Can this be used for TDD?

Absolutely. It includes specific patterns for Red-Green-Refactor cycles, ensuring that regression tests actually failed before the fix was applied.

How does it improve code quality?

It eliminates 'false positives' and premature success claims by requiring fresh evidence (like test logs or exit codes) before allowing a task to proceed to completion.

Verification Before Completion

Name: Verification Before Completion
Author: ed3dai

byed3dai

•

安全与测试

Enforces rigorous evidence-based verification before any task completion claims, commits, or pull requests.

This skill acts as a mandatory quality gate for Claude, preventing it from making premature or unverified claims of success. It establishes an 'Iron Law' where every assertion—whether a bug fix, a successful build, or passing tests—must be supported by fresh, explicitly executed command output. By identifying and blocking rationalizations like 'it should work' or 'I'm confident,' it ensures that only truly validated work moves forward, significantly reducing regressions and ensuring that the developer can trust every success report.

主要功能

01Mandatory gate function requiring fresh command execution before any success claim

02Structured Red-Green verification for regression tests and TDD workflows

03Detection and prevention of 'soft' language like 'should pass' or 'seems fixed'

04Comprehensive validation checklists covering builds, linters, tests, and requirements

05Independent verification requirements for delegated agent tasks and automated changes

0645 GitHub stars

使用场景

01Preventing broken code from being committed by requiring a full clean build and test run

02Ensuring bug fixes are validated against the original symptoms before being marked as resolved

03Validating that complex project requirements are met through a line-by-line checklist verification

What are Skills?·How to Install

Install with 🐟 Skill.Fish

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

For use in Claude.ai and ChatGPT

Download Skill