How does it handle agent delegation?

It requires the primary agent to independently verify the changes made by sub-agents rather than blindly trusting their internal success reports.

Does this skill slow down the development process?

While it adds a mandatory verification step, it saves significant time by preventing rework and catching errors before they reach the repository.

What does this skill actually prevent?

It prevents the AI from claiming a task is finished based on assumptions or partial checks, requiring actual command output as proof of completion.

Why is 'evidence before claims' so important?

It builds trust and prevents the deployment of broken code or incomplete features that might 'seem' to work but fail under actual execution.

Can I use this with any testing framework?

Yes, it is framework-agnostic and works with any command-line verification tool including Jest, Pytest, linters, and compilers.

Verified Completion Guard

Name: Verified Completion Guard
Author: claudiodearaujo

byclaudiodearaujo

Security & Testing

Enforces a strict evidence-first protocol that requires running verification commands before claiming any task is complete or passing.

About

Verified Completion Guard is a high-integrity skill designed to eliminate premature success claims and 'hallucinated' progress in AI-driven development. It acts as a cognitive gatekeeper by mandating that Claude identifies, executes, and analyzes actual command output—such as test results, build logs, or linter reports—before making any assertions of success. By adhering to the 'Iron Law' of evidence before claims, this skill ensures that code changes are truly functional, requirements are fully met, and regression tests are validated through a proper red-green cycle, ultimately building a foundation of trust and reliability in the development workflow.

Key Features

Mandates fresh command execution for every success assertion
Standardizes verification patterns for tests, builds, and linters
Enforces rigorous red-green TDD cycles for bug fixes
0 GitHub stars
Blocks non-committal language like 'should' or 'probably'
Requires independent verification of delegated agent tasks

Use Cases

Ensuring production readiness by confirming zero-error linter and build outputs
Preventing premature task closure during complex multi-step refactoring
Validating bug fixes with verified regression evidence before creating pull requests

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add claudiodearaujo/izacenter verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill

GitHub