What happens if a request is ambiguous?

The skill identifies the specific ambiguity and uses the 'AskUserQuestion' tool to seek clarification until the implementation confidence threshold is met.

What is the phase-2-refine skill?

It is a specialized component of the Plan Marshall workflow that analyzes and clarifies coding requests to ensure high-quality, predictable AI outputs.

Does it understand my project's specific technologies?

Yes, it loads a pre-computed architecture context including technology stacks, module purposes, and dependency graphs to ensure the request is technically valid for your specific project.

Can I run this skill manually?

No, this skill is configured as user-invocable false, meaning it is designed to be triggered automatically as part of a structured AI orchestration sequence.

How does it validate requests?

The skill evaluates requirements across five quality dimensions: correctness, completeness, consistency, non-duplication, and ambiguity, using the project's architecture as a reference.

AI Request Refinement & Analysis

Name: AI Request Refinement & Analysis
Author: cuioss

bycuioss

•

Productivity & Workflow

Refines and validates AI coding requests through iterative analysis of technical requirements and project architecture.

The Phase 2 Refine skill is a critical orchestration component within the Plan Marshall framework that ensures coding requests meet a high confidence threshold before implementation. It systematically analyzes requests for correctness, completeness, consistency, and ambiguity while mapping them against the existing codebase architecture. By identifying potential conflicts or missing information early, it prevents wasted compute and ensures that AI-generated solutions are technically feasible and perfectly aligned with the project's established modules and patterns.

Key Features

01Intelligent module mapping to determine implementation scope

02Configurable compatibility strategies for breaking changes or deprecations

03Multi-dimensional quality checks for correctness, consistency, and ambiguity

04Iterative requirement clarification until confidence thresholds are reached

051 GitHub stars

06Automated architecture-aware validation against project modules and technologies

Use Cases

01Resolving ambiguous requirements through automated Q-Gate findings

02Ensuring technical feasibility of tasks before generating implementation outlines

03Validating complex feature requests against existing project architecture

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cuioss/plan-marshall phase-2-refine

For use in Claude.ai and ChatGPT

Download Skill