What are the 'Red Flags' this skill monitors?

It flags phrases like 'should work,' 'seems to,' or 'probably,' as well as expressions of satisfaction like 'Great!' or 'Done!' before verification evidence is produced.

How does it prevent 'hallucinated' success in AI coding?

It implements a 'Gate Function' that requires the model to identify a proof command, run it, read the full output, and confirm the exit code before it is allowed to use completion language.

Can this skill be used for regression testing?

Yes, it includes specific patterns for Red-Green cycles, requiring the model to prove the failure exists before proving the fix works.

Is this skill compatible with CI/CD workflows?

Absolutely. It mirrors best practices for CI/CD by treating build exit codes and linter outputs as the only valid indicators of a successful task state.

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

Its primary purpose is to ensure that Claude never claims a task is finished or a bug is fixed without providing fresh, empirical evidence from a command execution.

Verification Before Completion

Name: Verification Before Completion
Author: carlopezzuto

bycarlopezzuto

0•

Security & Testing

Enforces a strict evidence-based workflow that requires fresh command output before any task can be claimed as complete.

The Verification Before Completion skill establishes a rigorous protocol for Claude Code to ensure high-integrity development. By implementing the 'Iron Law' of evidence-before-assertions, it prevents the model from making premature claims of success or satisfaction. This skill forces a systematic 'Gate Function' where every fix, build, or test must be verified with fresh execution data and exit codes before proceeding. It is specifically designed to eliminate common AI failures such as hallucinated fixes, partial verifications, and reliance on 'should work' logic, making it an essential tool for developers who prioritize reliability and truth over speed.

Key Features

01Mandatory gate function for all success and completion claims

02Independent verification protocols for delegated agent tasks

03Evidence-based reporting requiring fresh command execution output

04Detection and prevention of 'rationalization' language (e.g., 'should work')

050 GitHub stars

06Red-green TDD cycle enforcement for validated regression testing

Use Cases

01Ensuring all unit and integration tests pass before committing code

02Validating full build success and linter cleanliness before PR creation

03Verifying bug fixes through documented reproduction and resolution cycles

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