Can I skip verification if the code looks correct?

No. The skill's 'Iron Law' states that confidence is not a substitute for evidence; only fresh command output can validate a claim of completion.

What are the common 'red flags' this skill identifies?

It flags language like 'should work', 'probably fixed', or 'seems to be passing' as insufficient, requiring actual logs instead.

Why should I use the Verification Before Completion skill?

It prevents 'false success' reports by forcing Claude to provide actual command output evidence before claiming a task is finished, building trust in AI-generated work.

How does it handle tasks delegated to other agents?

It mandates that the primary agent verify the VCS diffs and actual state independently rather than trusting a 'success' report from a sub-agent.

What types of verification does this skill support?

It is designed to handle unit tests, linter outputs, build logs, manual regression tests, and requirement checklist validations.

Strict Verification Before Completion

Name: Strict Verification Before Completion
Author: lifangda

bylifangda

•

보안 및 테스팅

Enforces a rigorous verification workflow requiring empirical evidence before any task is marked as complete or successful.

This skill transforms how Claude handles task finalization by mandating an 'evidence-first' approach, preventing premature or inaccurate claims of success. It establishes a non-negotiable gate function where every assertion—whether a bug fix, a passing test, or a successful build—must be backed by fresh command output and objectively verified results. Ideal for high-stakes development environments where accuracy is critical, it eliminates assumptions, 'should-works', and incomplete verification, ensuring that every PR or commit meets the highest standards of integrity and functional correctness.

주요 기능

01Red-green regression testing validation (TDD)

02Anti-rationalization checks for 'should-work' assumptions

0316 GitHub stars

04Mandatory fresh verification command execution

05Empirical evidence checking for all success claims

06Independent verification of agent-delegated tasks

사용 사례

01Ensuring full build and lint compliance before pushing to a repository.

02Validating complex requirements against a line-by-line checklist with proof of execution.

03Confirming bug fixes with verified regression tests before opening a pull request.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lifangda/claude-plugins verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill