What triggers the verification gate function?

The skill activates whenever a task is near completion, specifically before commits, pull requests, or any communication suggesting a task is finished.

Can I trust the AI's success reports without this skill?

Without this skill, an AI might assume success based on code changes alone; this skill requires fresh command-line evidence as proof of correctness.

How does it improve regression testing?

It mandates a Red-Green cycle: the AI must prove the failure exists by running a test, apply the fix, and then provide fresh output confirming the test now passes.

Does it support linter and build verification?

Yes, it distinguishes between linter passes and full build successes, requiring the AI to execute and confirm the exit codes for each specific tool.

Why use the Verification Before Completion skill?

It prevents the AI from making premature or inaccurate claims about task success, ensuring all code is empirically functional and tested before you review it.

Evidence-Based Task Verification

Name: Evidence-Based Task Verification
Author: carlopezzuto

bycarlopezzuto

0•

보안 및 테스팅

Enforces a rigorous protocol requiring empirical evidence and command output before any task is marked as complete or successful.

The Verification Before Completion skill establishes a mandatory 'Iron Law' for Claude, preventing it from making premature or unsubstantiated claims of success. By implementing a gate function that requires identifying, running, and reading the output of specific verification commands, the skill ensures that code changes, bug fixes, and builds are actually functional before a PR is created or work is handed off. This approach eliminates common pitfalls like relying on 'false confidence' or assuming a linter pass equates to a successful build, fostering a higher level of trust and technical accuracy in AI-assisted development.

주요 기능

01Systematic verification of exit codes and failure counts

02Red-Green regression testing enforcement to validate bug fixes

03Independent validation of agent-delegated work and VCS diffs

04Mandatory execution of verification commands before success claims

05Explicit prohibition of vague language like 'should work' or 'seems to'

060 GitHub stars

사용 사례

01Ensuring full build and linter compliance before committing changes to a repository

02Validating that a bug is fully resolved by re-testing the original symptom before claiming a fix

03Verifying that all project requirements are met through a line-by-line evidence checklist

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add carlopezzuto/agents verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill