spawn_skill

Overview

This skill generates production-quality, PhD-level expert agent prompts for Claude Code. It produces comprehensive 500–1000 line Markdown agents with YAML frontmatter, real code examples, full configs, patterns, and authoritative references. Use the skill to create project-scoped or global user-scoped agents at scale, or to analyze an architecture and recommend agents automatically.

How this skill works

The skill offers three modes: Single Agent, Batch Generation, and Architecture Analysis. It asks a single multi-question form (mode, scope, output, platform(s), focus areas), then constructs a complete agent Markdown file following a 10-part template (core concepts, patterns, configs, tests, security, etc.). It can save files to project or global agent directories or return clipboard-ready content.

When to use it

• You need a complete, canonical expert agent for a specific technology (500–1000 lines).
• You want to generate many agents in one run from a list of platforms.
• You want the tool to parse an architecture and suggest which agents to create.
• You need agents with production-ready configs, tests, and canonical references.
• You prefer consistent agent structure and automated file placement (project vs global).

Best practices

• Answer the multi-question form in one submission so the skill can generate the correct UI and outputs.
• Pick project scope for team-shared agents (version controlled) and global scope for personal reusable agents.
• Limit tools in YAML frontmatter to only the necessary tools to reduce attack surface and focus behavior.
• Provide explicit focus areas (caching, pub/sub, CI/CD) to tailor patterns and code examples.
• Review generated agents immediately, test with sample tasks, and iterate—treat output as a curated starting point, not a final deployment.

Example use cases

• Create a Redis expert agent with caching, pub/sub, and scaling patterns and save it globally for all projects.
• Batch-generate agents for PostgreSQL, RabbitMQ, and AWS S3 as project-level files for a microservices repo.
• Analyze an e-commerce architecture and create selected agents for Next.js, Node API, Postgres, Stripe, and Redis.
• Produce a clipboard-ready expert prompt for a new framework so you can paste and refine in the Claude Code UI.