Can I use this for pull request preparation?

Yes, the skill is designed to be applied always before committing, creating PRs, or moving to the next task to maintain a high trust level and project integrity.

What counts as valid verification evidence?

Valid evidence includes fresh command output showing zero test failures, linter reports with zero errors, and build commands returning an exit code of 0.

How does it handle bug fixes?

The skill enforces a red-green cycle: the AI must run the test to see it fail, apply the fix, and then run it again to see it pass, proving the fix works.

Why is this skill necessary for AI development?

It prevents the AI from making premature or false claims of success, ensuring that every code change is backed by fresh, empirical evidence from tests or builds.

Verification Before Completion

Name: Verification Before Completion
Author: tianyuanztm

bytianyuanztm

0•

Security & Testing

Enforces rigorous, evidence-based verification through command execution before any work is claimed as complete or submitted.

This skill establishes a critical gate function for AI-driven development by mandating fresh, empirical evidence before any success claims are made. It prevents the common pitfall of 'hallucinated completion' by requiring Claude to execute verification commands—such as tests, linters, or builds—and analyze the actual output and exit codes before committing code or creating pull requests. By prioritizing 'evidence before assertions,' it builds high-trust workflows, prevents the shipping of undefined functions or missing requirements, and ensures that the agent's work genuinely meets the project's quality standards.

Key Features

01Implements a structured Gate Function workflow (Identify, Run, Read, Verify, Claim)

020 GitHub stars

03Enforces red-green cycles for regression testing to prove bug fixes are effective

04Mandates fresh execution of verification commands before making any success claims

05Identifies and stops usage of 'red flag' language like 'should' or 'probably'

06Prevents blind trust in delegated agent reports by requiring independent VCS diff checks

Use Cases

01Cross-referencing task completion against original requirements using a line-by-line checklist

02Automating pre-commit checks to ensure tests and linters pass with zero failures

03Verifying bug fixes through red-green cycles to ensure the original symptom is resolved

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tianyuanztm/myplan verification-before-completion

For use in Claude.ai and ChatGPT

Key Features

01Implements a structured Gate Function workflow (Identify, Run, Read, Verify, Claim)

020 GitHub stars

03Enforces red-green cycles for regression testing to prove bug fixes are effective

04Mandates fresh execution of verification commands before making any success claims

05Identifies and stops usage of 'red flag' language like 'should' or 'probably'

06Prevents blind trust in delegated agent reports by requiring independent VCS diff checks

Use Cases

01Cross-referencing task completion against original requirements using a line-by-line checklist

02Automating pre-commit checks to ensure tests and linters pass with zero failures

03Verifying bug fixes through red-green cycles to ensure the original symptom is resolved