What is the primary goal of the Verification Before Completion skill?

The primary goal is to ensure honesty and accuracy by requiring Claude to run verification commands and see successful output before claiming any task is finished.

Does this skill work with linter and build tools?

Yes, it requires explicit execution of linters and compilers, ensuring that a 'clean' status is backed by an exit code of 0 and zero reported errors.

Can this skill be used for regression testing?

Yes, it specifically enforces a red-green verification cycle where a test must fail after removing a fix and pass after restoring it to prove the fix works.

What are considered 'Red Flags' by this skill?

Red flags include language like 'should work,' 'probably fixed,' or 'seems correct,' as well as expressing satisfaction before fresh verification evidence is produced.

How does this skill prevent 'hallucinated' success?

It implements a 'Gate Function' that forces the AI to identify a proof command, run it, read the full output, and verify the exit code before making any assertions.

Strict Output Verification

Name: Strict Output Verification
Author: vgeshel

byvgeshel

•

セキュリティとテスト

Enforces an evidence-based workflow that requires running verification commands before claiming task completion or creating pull requests.

Strict Output Verification is a quality assurance skill designed to eliminate premature success claims by establishing a non-negotiable protocol for Claude Code. It mandates that every assertion—whether it is a passing test suite, a clean linter report, or a successful build—must be supported by fresh, visible command output rather than assumptions or previous state. By identifying 'red flag' language and enforcing a rigorous gate function, this skill ensures higher code reliability, prevents the submission of broken features, and maintains a high level of trust between the AI and the developer.

主な機能

01Provides a standardized 'Gate Function' for state verification

02Mandates fresh command execution for every success claim

03Identifies and prevents rationalization patterns and vague language

04Validates agent delegation results through VCS diff checks

05Enforces TDD red-green cycles for regression testing

061 GitHub stars

ユースケース

01Ensuring full test suite compliance before opening a pull request

02Validating project requirements against a line-by-line checklist before handoff

03Verifying bug fixes through red-green cycles to confirm regression prevention

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add vgeshel/local-skills verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill