When should I use the 'VF' task format?

The Verification (VF) task should be added to your development specs after PR creation to ensure the original issue is explicitly verified as resolved before merging.

What is the primary purpose of the Reality Verification skill?

The skill ensures that any code fix is preceded by a successful reproduction of the failure and followed by a documented verification that the specific failure is resolved.

Which tools does the Reality Verification skill support?

It maps commands for common tools including GitHub CLI (gh), pnpm/npm/yarn, TypeScript (tsc), Playwright for E2E, and standard fetch for API testing.

Can this skill detect low-quality tests?

Yes, it includes mock-quality checks that flag 'red flags' such as tests that only assert mock interactions without verifying real module behavior.

How does it track the progress of a fix?

It uses a structured format in a .progress.md file to record the 'BEFORE' state (failure output and timestamp) and the 'AFTER' state (success output and comparison).

Reality Verification

Name: Reality Verification
Author: tzachbon

bytzachbon

•

131

•

Security & Testing

Enforces a rigorous workflow for reproducing failures before implementation and verifying resolutions after code changes.

The Reality Verification skill enhances Claude Code by implementing a strict 'repro-first' methodology for bug fixes and feature development. It maps user goals to specific diagnostic commands—such as CI logs, test suites, and linter checks—to establish a documented 'BEFORE' state of failure. By requiring comparative documentation in progress files and performing quality checks to detect mock-only test anti-patterns, this skill ensures that fixes are genuine, resolutions are verifiable, and the development loop remains grounded in empirical evidence rather than assumptions.

Key Features

01Automated goal detection to distinguish between bug fixes and new feature requests

02131 GitHub stars

03Standardized VF (Verification) task generation for spec-driven development workflows

04Mock-quality checks to identify and flag weak or deceptive test patterns

05Structured BEFORE/AFTER state documentation for clear audit trails in .progress.md

06Dynamic command mapping for CI logs, unit tests, linting, and E2E verification

Use Cases

01Providing a verifiable comparison between a failing error message and a passing test result

02Ensuring a bug reported in a remote CI pipeline is successfully reproduced locally before starting work

03Validating that a complex refactor hasn't introduced regressions by re-running the original reproduction steps

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tzachbon/smart-ralph reality-verification

For use in Claude.ai and ChatGPT

Download Skill