What is the primary purpose of the Verification Guard skill?

It ensures Claude never claims a task is finished without first running a command (like a test or build) that proves the claim is true.

Can this skill be used with any programming language?

Yes, it is language-agnostic and applies to any workflow involving command-line verification tools such as npm, pytest, go test, or compilers.

Does it work when Claude is using other agents?

Yes, it requires Claude to independently verify the changes made by sub-agents rather than blindly trusting their internal success reports.

Why is 'Evidence before claims' the core principle?

Claiming work is done without running a fresh check can lead to shipping broken code; requiring evidence ensures honesty and technical accuracy in every response.

How does this skill prevent AI hallucinations in code success?

It strictly forbids the use of speculative language like 'should' or 'looks good' and requires the AI to read and confirm the actual exit code of verification commands.

Verification Guard

Name: Verification Guard
Author: BbgnsurfTech

byBbgnsurfTech

•

Security & Testing

Enforces rigorous evidence-based reporting by requiring fresh verification results before any task is claimed as complete.

The Verification Guard skill acts as a critical quality gate for Claude, preventing premature or inaccurate 'done' claims by mandating the execution of actual verification commands. It moves Claude's workflow from an assumption-based model to an evidence-based one, requiring successful test runs, build completions, or linter passes before a status can be reported. By strictly forbidding words like 'should' or 'probably,' this skill ensures that every assertion of success is backed by raw output and exit codes, significantly reducing regressions and building high levels of trust in AI-driven development.

Key Features

013 GitHub stars

02Analyzes exit codes and raw logs to confirm actual success

03Validates agent delegation results through independent diff analysis

04Detects and blocks vague language and premature satisfaction statements

05Enforces a strict Red-Green cycle for regression testing

06Mandates fresh command execution for every completion claim

Use Cases

01Ensuring strict TDD compliance during complex bug fixing and refactoring

02Preventing broken code or failing tests from being committed to a repository

03Verifying that all requirements are met through line-by-line checklist validation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bbgnsurftech/claude-skills-collection verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill