Can Claude skip verification if the change is small?

No, the skill's 'Iron Law' prohibits any exceptions, ensuring that even minor changes are verified to prevent unexpected regressions.

Why is 'evidence before assertions' important?

It prevents 'hallucinated' success where the AI assumes code works without actually testing the execution path, maintaining trust between the AI and the developer.

What does this skill actually do?

It forces Claude to run a verification command and show the output before it is allowed to say a task is finished, ensuring every success claim is backed by data.

Does this work with any test runner?

Yes, the skill is command-agnostic and applies to any CLI-based verification tool including Jest, Pytest, ESLint, or custom build scripts.

Verification Before Completion

Name: Verification Before Completion
Author: seanGSISG

byseanGSISG

セキュリティとテスト

Mandates the execution of verification commands and evidence gathering before any task is marked as complete.

概要

This skill enforces a rigorous 'evidence-first' workflow for Claude, preventing it from claiming success or completion without providing fresh verification data. By requiring a strict gate function—identifying the proof command, executing it, and analyzing the full output—it eliminates common failures like 'should-work' assumptions and unverified bug fixes. It is essential for maintaining high code quality and trust, ensuring that every claim of success is backed by recent test results, linter outputs, or build logs rather than mere confidence.

主な機能

Implements a 5-step gate function (Identify, Run, Read, Verify, Claim)
Identifies and blocks 'red flag' language like 'should' or 'probably'
Supports TDD workflows with required Red-Green-Refactor verification
0 GitHub stars
Mandates fresh verification evidence for every success claim
Prevents blind trust in delegated agent reports and partial linter passes

ユースケース

Ensuring unit and integration tests pass before opening a Pull Request
Verifying bug fixes through documented Red-Green regression cycles
Confirming build and lint success in complex CI/CD environments

概要

主な機能

Implements a 5-step gate function (Identify, Run, Read, Verify, Claim)
Identifies and blocks 'red flag' language like 'should' or 'probably'
Supports TDD workflows with required Red-Green-Refactor verification
0 GitHub stars
Mandates fresh verification evidence for every success claim
Prevents blind trust in delegated agent reports and partial linter passes

ユースケース

Ensuring unit and integration tests pass before opening a Pull Request
Verifying bug fixes through documented Red-Green regression cycles
Confirming build and lint success in complex CI/CD environments