How does this improve the developer-AI workflow?

It builds trust by ensuring that when the AI reports a task is finished, it has actually been validated, reducing time spent on manual re-checking and rework.

What counts as valid verification evidence?

Fresh terminal output from test runners, linters, or build commands that confirms a successful exit code and zero failures.

Can this skill be used for regression testing?

Yes, the skill explicitly enforces a red-green verification cycle to ensure regression tests are functional and correctly identify the fix.

Why is the Verification Before Completion skill necessary?

It prevents AI agents from making premature or inaccurate claims about task completion by requiring empirical evidence from command outputs before any success is reported.

Verification Before Completion

Name: Verification Before Completion
Author: RockwallJMC

byRockwallJMC

•

安全与测试

Enforces a mandatory evidence-based protocol requiring terminal command verification before any task is marked as complete or successful.

This skill implements a rigorous 'Iron Law' of verification for AI agents, preventing premature claims of success or completion. It establishes a strict gate function where Claude must identify, execute, and interpret the output of verification commands—such as test suites, linters, or build scripts—before expressing satisfaction or committing changes. By prioritizing empirical evidence over assumptions, it eliminates common failures like 'should work' syndrome and ensures that all code changes are validated against actual results, building significantly higher trust between the AI and human developers.

主要功能

01Independent verification protocols for agent-delegated tasks and VCS diffs

021 GitHub stars

03Red-green TDD cycle verification for robust regression testing

04Standardized gate function for identifying and executing validation commands

05Strict anti-rationalization guidelines to prevent biased success reports

06Mandatory 'Iron Law' requiring fresh command output before any success claim

使用场景

01Preventing broken or unverified code from being committed to a repository

02Validating bug fixes by verifying the resolution of the original symptom with evidence

03Ensuring unit and integration tests pass with 0 failures before task handoff

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rockwalljmc/piercedesk6 VERIFY-BEFORE-COMPLETE

For use in Claude.ai and ChatGPT

Download Skill