When should this skill be applied?

It should be applied before any variation of success claims, expressions of satisfaction, commits, PR creations, or moving to the next task.

What is the 'Iron Law' in this skill?

The Iron Law states that no completion claims can be made without fresh verification evidence. If a verification command hasn't been run in the current context, the task cannot be claimed as passing.

Does this skill work with TDD workflows?

Yes, it specifically includes patterns for Red-Green verification, requiring proof that a test fails without a fix before confirming it passes with the fix.

How does this prevent AI 'hallucinations' regarding code success?

It forces the agent to execute specific commands and read the full output before making any positive assertions, moving from 'assuming' to 'proving' results.

Verification Before Completion

Name: Verification Before Completion
Author: franroa

byfranroa

0•

セキュリティとテスト

Enforces rigorous, evidence-based verification protocols before any task is claimed as complete or submitted for review.

This skill acts as a mandatory quality gate for AI agents, ensuring that no status claims, commits, or pull requests are made without fresh, documented proof of success. By implementing a strict 'Evidence Before Assertions' policy, it eliminates common failures such as 'hallucinated success,' partial linter checks, and unverified bug fixes. It is an essential tool for maintaining high code quality standards and ensuring that every code change is backed by actual command output, test results, and requirement checklists.

主な機能

01Anti-rationalization guardrails to prevent common 'should work' assumptions

02Mandatory 'Iron Law' requiring fresh verification evidence for all claims

03Systematic 5-step Gate Function: Identify, Run, Read, Verify, and Claim

04Integrated Red-Green TDD cycle verification for regression testing

050 GitHub stars

06Cross-functional validation for builds, linters, tests, and requirement checklists

ユースケース

01Preventing premature PR creation by requiring a full successful test suite execution

02Auditing delegated agent tasks by inspecting VCS diffs and running independent builds

03Verifying bug fixes by proving the original symptom is gone through fresh output

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add franroa/chezmoi verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill