ck:skill-creator

Create and improve Claude skills through eval-driven iteration.

Updated Apr 18, 2026
One-click install
npx skills add https://github.com/lengo0951/pms --skill ck-skill-creator-lengo0951
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ck:skill-creator
Source: https://github.com/lengo0951/pms/tree/main/.claude/skills/skill-creator
Command: npx skills add https://github.com/lengo0951/pms --skill ck-skill-creator-lengo0951

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires python, scripts, agents, evals, metadata, and includes scripts (resource) and agents (resource) and evals (resource) and metadata (resource) components.

What problem does it solve?

This Skill helps users create or update Claude skills by guiding them through the process of eval-driven iteration, making it easier to develop effective and optimized skills.

Core Features & Use Cases

  • Skill Creation: Walks users through the process of creating new Claude skills.
  • Skill Update: Allows users to update existing skills with new descriptions, scripts, and references.
  • Eval-Driven Iteration: Provides a framework for testing and optimizing skills through evaluation and feedback loops.
  • Use Case: A user wants to create a new skill for processing CSV files. They can use this Skill to guide them through the process, from capturing intent and planning to testing and iterating on the skill.

Quick Start

To create a new skill, run the following command: scripts/init_skill.py <name> --path <dir>.

Frequently Asked Questions about ck:skill-creator

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

FAQPage Schema
How do I create new Claude skills using eval-driven iteration?

You can create new Claude skills by running the initialization script to capture intent, plan the structure, and then use evaluation feedback loops to iteratively test and optimize the skill descriptions and logic.

Can I update and benchmark existing Claude skills?

Yes, you can update existing Claude skills by adding new descriptions, scripts, and references, and then benchmarking them through parallel testing and grading to validate performance improvements.

Do I need Python to use the Claude skill creation and optimization process?

Yes, Python is required along with specific Claude-related libraries to run the initialization, evaluation, and packaging scripts that drive the skill creation and benchmarking workflow.

How does evaluation and feedback work when optimizing a skill?

Optimization uses an evaluation framework that supports parallel testing, automated grading, and human-in-the-loop feedback to validate skill behavior and guide iterative improvements.

What is the best way to structure a skill for processing specific file formats like CSV?

The best way is to use the skill initialization script to capture your specific intent, plan the required scripts and agents, and then iteratively refine the skill through structured benchmarking.

Are there limitations when using automated grading for skill validation?

Automated grading validates skill behavior during benchmarking, but the framework also supports human-in-the-loop feedback to catch edge cases and ensure the skill meets complex intent requirements.