phase-2-refine

Iteratively evaluate and refine plan requests until a confidence threshold is reached.

4|Updated Jan 11, 2026
One-click install
npx skills add https://github.com/cuioss/plan-marshall --skill phase-2-refine
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: phase-2-refine
Source: https://github.com/cuioss/plan-marshall/tree/main/marketplace/bundles/plan-marshall/skills/phase-2-refine
Command: npx skills add https://github.com/cuioss/plan-marshall --skill phase-2-refine

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Phase 2 refine provides an iterative workflow to clarify and validate plan requests so subsequent phase outputs are accurate and actionable.

Core Features & Use Cases

  • Iterative clarification loop until requirements are unambiguous
  • Architecture context loading and module mapping to ensure feasibility
  • Confidence scoring and decision logging to guide progression or rework
  • Support for user clarifications and request updates
  • Seamless transition to phase 3 outline when the threshold is reached

Quick Start

Initiate the phase-2 refine loop for a plan to drive clarification until the confidence threshold is reached.

Frequently Asked Questions about phase-2-refine

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I clarify ambiguous plan requirements before starting architecture work?

To clarify ambiguous plan requirements, an iterative refinement loop assesses correctness, completeness, consistency, and ambiguity against architecture context. This structured evaluation maps findings to modules and logs decisions until a confidence threshold is reached.

What is the best way to validate a planning request for completeness and consistency?

Validating a planning request involves loading architecture context and configuration to evaluate the request iteratively. It checks for correctness, completeness, consistency, duplication, and ambiguity, mapping any findings to specific modules for resolution.

How does iterative requirement refinement work with a confidence threshold?

Iterative requirement refinement works by cycling through evaluations of user clarifications and request updates. It logs decisions and continues assessing the plan until the predefined confidence threshold is met or surfaces findings for manual review.

Can I manually review findings when requirement confidence threshold is not reached?

Yes, when the confidence threshold is not reached through user clarifications, the workflow surfaces the accumulated findings for manual review. This ensures unresolved ambiguities or inconsistencies are addressed before proceeding.

When do I need an iterative clarification loop for my project workflow?

You need an iterative clarification loop when your plan requests require validation against architecture context to ensure feasibility. It is essential when subsequent phase outputs must be accurate, actionable, and free of duplication or ambiguity.

Does phase-2-refine support transitioning to the next planning phase automatically?

Yes, phase-2-refine supports transitioning to phase 3 outline automatically. Once the iterative evaluation reaches the predefined confidence threshold, it seamlessly progresses the workflow to the next phase.