skill-creator

Create and iteratively improve Agent Skills through structured evaluations.

Updated Apr 13, 2026
One-click install
npx skills add https://github.com/bermudi/skills --skill skill-creator-bermudi
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/bermudi/skills/tree/main/skill-creator
Command: npx skills add https://github.com/bermudi/skills --skill skill-creator-bermudi

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

The Skill Creator provides a structured, eval-driven workflow to design, test, and iteratively improve Agent Skills, reducing time-to-value and increasing triggering reliability.

Core Features & Use Cases

  • Create and iteratively improve Agent Skills using a repeatable process: draft, run evaluative prompts, measure, and refine.
  • Run end-to-end evals with train/test splits, baselines, and benchmarking to compare iterations.
  • Manage test prompts, assertions, and output quality with progressive disclosure and on-demand references.

Quick Start

Create a new skill by drafting its SKILL.md, run iterative evals, and optimize the description for triggering accuracy.

Frequently Asked Questions about skill-creator

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

FAQPage Schema
How do I create and iteratively improve agent skills for better triggering accuracy?

To create agent skills, you draft a structured SKILL.md prompt, run iterative evaluative test prompts, measure the results against baselines, and refine the description to increase triggering reliability.

What is an eval-driven workflow for skill creation and why is it needed?

An eval-driven workflow for skill creation is a repeatable process of drafting, running end-to-end evaluations, and refining. It is needed to reduce time-to-value and ensure reliable agent skill triggering.

How do I run end-to-end evals with train and test splits for agent skills?

You run end-to-end evals for agent skills by using the provided scripts to manage test prompts, execute train/test splits, establish baselines, and benchmark current iterations against previous versions.

Can I benchmark skill iterations to compare output quality and triggering reliability?

Yes, you can benchmark skill iterations to compare output quality. The workflow supports measuring test results against established baselines to evaluate progressive improvements during skill creation.

What's the best way to structure a SKILL.md prompt for rapid discovery and reuse?

The best way to structure a SKILL.md prompt for rapid discovery is to write a concise, structured description of the skill's core purpose, enabling the system to quickly find and reuse the skill.