Can I skip verification if I am confident the fix works?

No, the core principle of this skill is that 'confidence is not evidence.' It enforces a non-negotiable gate function to maintain high trust and reliability in the development process.

Why use the Verification Guard skill?

It prevents Claude from making false or premature claims about code status by requiring real-time proof from test runs, linters, and build commands.

Does it work with different programming languages?

Yes, it is language-agnostic. It works by enforcing the process of running whatever verification commands (such as npm test, pytest, or go test) are relevant to your specific project.

How does this help with pull requests?

It ensures that no PR is suggested or created until every change has been validated with fresh evidence, significantly reducing review cycles and broken builds.

Verification Guard

Name: Verification Guard
Author: abudhahir

byabudhahir

•

セキュリティとテスト

Enforces rigorous evidence-based validation of code changes and test results before any task is marked as complete.

概要

Verification Guard is a quality assurance skill for Claude Code that implements a strict 'evidence-before-assertions' protocol. It prevents premature completion claims by requiring the AI to execute fresh verification commands, analyze full outputs, and confirm success with concrete evidence before committing code or creating pull requests. By eliminating 'false positives' and 'it should work' assumptions, this skill ensures that every bug fix, feature, and refactor is actually functional and meets all project requirements before the task is handed off.

主な機能

Prevents rationalization and 'false confidence' during the final stages of a task
Validates build, linter, and test outputs against actual exit codes and failure counts
Requires line-by-line requirement checklist verification for complex features
Mandates fresh command execution before any success or completion claims
1 GitHub stars
Implements a strict Red-Green verification cycle for regression testing

ユースケース

Validating that large-scale refactors haven't introduced regressions in the build process
Ensuring bug fixes actually resolve the original symptom through verified test runs
Confirming agent-delegated tasks meet requirements by inspecting diffs and running independent checks

概要

主な機能

Prevents rationalization and 'false confidence' during the final stages of a task
Validates build, linter, and test outputs against actual exit codes and failure counts
Requires line-by-line requirement checklist verification for complex features
Mandates fresh command execution before any success or completion claims
1 GitHub stars
Implements a strict Red-Green verification cycle for regression testing

ユースケース

Validating that large-scale refactors haven't introduced regressions in the build process
Ensuring bug fixes actually resolve the original symptom through verified test runs
Confirming agent-delegated tasks meet requirements by inspecting diffs and running independent checks