What is the primary purpose of the Verify Workflow skill?

It is designed to validate the outputs of AI-orchestrated coding tasks, ensuring they meet structural and qualitative standards across multiple development phases from planning to execution.

Which phases of the development lifecycle does this skill verify?

It supports a 6-phase model including Initialization (init), Refinement (refine), Outlining (outline), Planning (plan), Execution (execute), and Finalization (finalize).

Can I create my own custom test cases for my specific project?

Yes, the skill includes a 'create' subcommand that allows you to interactively define new test cases, version-control them, and establish 'golden' expected artifacts for future verification.

How does the hybrid assessment system work?

The skill uses Python scripts for deterministic structural checks (like file existence and schema validation) and LLM-as-judge for qualitative analysis (like evaluating logic or adherence to specific criteria).

Workflow Verifier

Name: Workflow Verifier
Author: cuioss

bycuioss

•

보안 및 테스팅

Validates AI-driven coding workflows using a hybrid system of automated scripts and LLM-as-judge assessments to ensure reliability.

The Verify Workflow skill provides a robust framework for validating the outputs of the plan-marshall orchestration layer across a standardized 6-phase model. By combining structural verification via Python scripts with qualitative assessment using LLM-as-judge patterns, it ensures that AI-driven coding tasks adhere to architectural standards, follow prescribed task structures, and produce valid artifacts. It is particularly useful for teams using Claude Code to automate complex engineering tasks where consistency and predictable outputs are critical for production-grade software development.

주요 기능

01Multi-phase verification covering initialization, refinement, outlining, planning, and execution.

02Automated structural checks for configuration files, task lists, and project deliverables.

031 GitHub stars

04Hybrid assessment engine combining deterministic Python scripts with qualitative LLM evaluations.

05Detailed scoring rubric (0-100) and failure taxonomy for root cause analysis.

06Interactive test case management including creation, listing, and automated execution against golden samples.

사용 사례

01Automating quality assurance for complex, multi-step code migrations and refactors.

02Regression testing AI assistant behavior against established baseline artifacts to prevent output drift.

03Validating that AI-generated project plans follow organizational architectural standards.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cuioss/plan-marshall verify-workflow

For use in Claude.ai and ChatGPT

Download Skill