How does the skill handle task dependencies?

It automatically scans the .prompts directory for existing research and plans, building a dependency graph to determine the correct execution order (Sequential or Parallel).

Why does every execution generate a SUMMARY.md file?

While the main outputs are XML-structured for AI parsing, the SUMMARY.md provides a substantive, human-readable overview of findings, decisions, and next steps.

What is a meta-prompt in this context?

A meta-prompt is a specialized instruction designed to generate output that serves as a structured input for another AI agent, enabling complex Claude-to-Claude communication chains.

Where are the generated prompts and artifacts stored?

All artifacts are organized in a local .prompts/ folder, using a numbered kebab-case naming convention (e.g., 001-auth-research) to track project progression.

What are the four primary purposes of the prompts created?

The skill categorizes prompts into: Research (gathering info), Plan (strategy/roadmap), Do (task execution/coding), and Refine (iterating on existing work).

Meta-Prompt Architect

Name: Meta-Prompt Architect
Author: shishiv

byshishiv

生産性とワークフロー

Orchestrates multi-stage Claude-to-Claude workflows through structured research, planning, and execution prompts.

概要

The Meta-Prompt Architect skill transforms Claude Code into a sophisticated workflow manager capable of handling complex, multi-step development tasks. It automates the generation of specialized prompts for Research, Planning, Implementation, and Refinement, ensuring that each stage produces structured XML outputs optimized for machine consumption. By maintaining a dedicated .prompts directory with versioned history and human-readable summaries, it provides a clear audit trail and dependency-aware execution engine for advanced AI-driven software engineering.

主な機能

Automated directory management and file versioning within .prompts/
Dependency-aware execution engine for sequential and parallel tasks
Multi-stage workflow orchestration (Research, Plan, Do, Refine)
Standardized XML output formatting for Claude-to-Claude communication
Human-centric SUMMARY.md generation for quick status oversight
0 GitHub stars

ユースケース

Creating structured prompt chains for automated code migrations and refactoring
Building complex features that require deep technical research before implementation
Documenting AI decision-making processes through persistent prompt provenance

概要

主な機能

Automated directory management and file versioning within .prompts/
Dependency-aware execution engine for sequential and parallel tasks
Multi-stage workflow orchestration (Research, Plan, Do, Refine)
Standardized XML output formatting for Claude-to-Claude communication
Human-centric SUMMARY.md generation for quick status oversight
0 GitHub stars

ユースケース

Creating structured prompt chains for automated code migrations and refactoring
Building complex features that require deep technical research before implementation
Documenting AI decision-making processes through persistent prompt provenance