new-skill

Create Claude Code Skills with YAML frontmatter and eval validation.

Updated Feb 28, 2026
One-click install
npx skills add https://github.com/cosmicdreams/claude-plugins --skill new-skill-cosmicdreams
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: new-skill
Source: https://github.com/cosmicdreams/claude-plugins/tree/main/admin/skills/new-skill
Command: npx skills add https://github.com/cosmicdreams/claude-plugins --skill new-skill-cosmicdreams

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires claude, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Creating or improving a Claude Code Skill is difficult because the description must trigger at the right times, while the body must provide clear, cold-start instructions in the correct plugin folder structure.

Core Features & Use Cases

  • Workflow-first skill creation: turns a vague request into a concrete skill plan by gathering triggers, expected outputs, and common failure modes.
  • Progressive disclosure structure: guides you to keep SKILL.md focused and move deeper detail into references/, scripts/, and assets/ with an explicit resource manifest.
  • Trigger-accuracy optimization loop: provides an eval infrastructure to test whether a Skill’s YAML description actually causes Claude to invoke it, and iterates on improvements.
  • Install-and-verify steps: covers reinstall verification and producing an eval record for shared learning across the system.

Quick Start

Use the new-skill workflow to create a new Skill by asking Claude to make a skill for "X" and writing SKILL.md to match the system’s progressive disclosure and trigger-accuracy requirements.

Frequently Asked Questions about new-skill

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

FAQPage Schema
How do I create a Claude Code Skill that triggers reliably?

To create a Claude Code Skill that triggers reliably, you must write a YAML frontmatter description that accurately matches user intents and structure the plugin folder with clear cold-start instructions inside SKILL.md.

What is progressive disclosure in Claude plugin structure?

Progressive disclosure in a Claude plugin structure involves keeping SKILL.md focused on core instructions while moving deeper detail into references, scripts, and assets folders using an explicit resource manifest.

How do I test whether my YAML description causes Claude to invoke a skill?

You test YAML trigger evaluation by using the bundled eval infrastructure to validate trigger behavior, which checks whether the skill description actually causes Claude to invoke it and allows for iterating improvements.

How do I structure skill folders for a new Claude Code plugin?

You structure skill folders for a Claude Code plugin by creating a root SKILL.md file and managing on-demand resources like references, scripts, and assets in separate subdirectories according to plugin conventions.

Why does my Claude Skill fail to trigger when users request related tasks?

A Claude Skill fails to trigger when its YAML frontmatter description does not accurately match user requests, requiring a trigger-accuracy optimization loop to refine the description and validate behavior.

Can I use automation scripts within a Claude Code Skill?

Yes, you can use automation scripts within a Claude Code Skill by placing them in the scripts directory and declaring them in the resource manifest to ensure they load correctly during execution.