Why does it use XML for communication?

XML provides a clear, structured format that Claude can parse with high reliability, ensuring that metadata, findings, and instructions are passed between prompts without losing context.

Can I review the AI's work without reading long prompt outputs?

Yes, every execution automatically generates a SUMMARY.md file designed for human scanning, providing a substantive one-liner, key findings, and proposed next steps.

What is a meta-prompt in the context of Claude Code?

A meta-prompt is a specialized prompt designed to guide Claude in generating other prompts or structured outputs that are specifically optimized for subsequent AI steps in a workflow.

How does this skill handle complex dependencies?

It automatically scans your prompts for file references to build a dependency graph, then executes them in the correct order—either sequentially for dependent steps or in parallel for independent ones.

Where are the generated prompts stored?

All prompts, archived versions, and outputs are stored in a standardized .prompts/ directory within your project to maintain clear version history and provenance.

Meta-Prompt Architect

Name: Meta-Prompt Architect
Author: jwilger

byjwilger

•

생산성 및 워크플로

Generates and executes multi-stage Claude-to-Claude prompt pipelines for complex research, planning, and implementation workflows.

Meta-Prompt Architect is a powerful workflow automation skill that transforms complex goals into structured, multi-step AI pipelines. By categorizing tasks into Research, Plan, Do, or Refine stages, it creates optimized prompts specifically designed for LLM-to-LLM communication using XML metadata. The skill automatically manages project organization within a dedicated prompts directory, ensuring clear provenance for every action. It bridges the gap between automated execution and human oversight by generating executive summaries alongside machine-readable outputs, making it indispensable for developers building autonomous agents or managing large-scale architectural changes.

주요 기능

01Intelligent dependency detection and execution (Sequential, Parallel, or DAG)

02Standardized project organization with automated folder and file versioning

037 GitHub stars

04XML-optimized structure for high-fidelity Claude-to-Claude communication

05Human-readable SUMMARY.md generation for rapid developer review

06Automated multi-stage workflow generation (Research, Plan, Implement, Refine)

사용 사례

01Iteratively refining project roadmaps based on newly discovered technical constraints

02Creating automated pipelines where one AI agent's research output informs another agent's plan

03Conducting deep-dive technical research and architectural planning before code implementation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jwilger/claude-code-plugins create-meta-prompts

For use in Claude.ai and ChatGPT

Download Skill