skills.rest
csgaokun
csgaokunCOMMUNITY
Explore all skills by this author →
GitHub

skill-creator

Draft a precise prompt summarizing a skill's purpose for vector store embedding.

Updated Aug 23, 2026
Search Tags:#claude#evaluation#benchmark#skill#prompt-engineering#creator
One-click install
npx skills add https://github.com/csgaokun/vscode_claude --skill skill-creator-csgaokun
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/csgaokun/vscode_claude/tree/main/.github/skills/skill-creator
Command: npx skills add https://github.com/csgaokun/vscode_claude --skill skill-creator-csgaokun

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

Drafting, refining, and validating Claude Skills is complex; this Skill unit provides a structured workflow to create new skills, iterate with evaluations, benchmark performance, and optimize triggering accuracy.

Core Features & Use Cases

  • Capture Intent: Define what the skill should enable Claude to do and how it should trigger.
  • SKILL.md authoring: Write a structured YAML frontmatter with name and description and craft the body to guide activation.
  • Iterative evaluation loop: Use built-in run_eval, run_loop, and improve_description tools to test triggers, measure performance, and iterate.
  • Benchmarking and review: Aggregate results with aggregate_benchmark and review outputs with the eval-review UI to compare with baselines and improvements.
  • Reference and tooling: Bundle scripts, references, and assets to support reproducible skills and on-demand resources.
  • Description optimization: Optionally improve triggering by running description optimization (scripts/improve_description.py) and previewing results.

Quick Start

Draft a new skill by writing SKILL.md with a clear name and description, then run the evaluation loop to refine prompts and measure improvements.

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 refine Claude Skills for prompt engineering?▼

Create and refine Claude Skills by authoring a structured SKILL.md with YAML frontmatter, then using an iterative evaluation loop to benchmark performance and optimize triggering accuracy.

What is the best way to evaluate and benchmark Claude prompt performance?▼

Evaluate and benchmark Claude prompt performance by running built-in evaluation loops that test triggers, measure improvements, and aggregate results to compare against baselines.

How do I improve the vector search discovery rate for my Claude Skill?▼

Improve vector search discovery by drafting a high-density prompt that concisely summarizes the skill's purpose and scope, and running description optimization scripts to refine triggering accuracy.

Can I use YAML frontmatter to define how a Claude Skill activates?▼

Yes, you define skill activation by writing a structured YAML frontmatter with name and description fields, then crafting the SKILL.md body to guide Claude's behavior during task execution.

What components are needed to build a reproducible Claude Skill?▼

Build reproducible Claude Skills by bundling scripts, references, and assets together, ensuring on-demand resources support the evaluation loop and description optimization processes.

Why are my Claude Skill triggers not activating correctly during benchmark testing?▼

Skill triggers may fail due to low-density descriptions; resolve this by running the improve_description script to optimize the prompt and previewing results before benchmarking again.

Related Skills

Explore More Software Engineering Skills→
Exia-thdExia-thd

skill-maker

Create, improve, and audit Forgewright skills with SKILL.md files and test cases.

Community
Advanced
randoneeringrandoneering

evaluate-cortex-agent

Evaluate Cortex Agents in Snowflake and review results in Snowsight.

Community
Advanced
Kingdaddy007Kingdaddy007
1

skill-creator

Define Skill metadata, test with evals, and iterate on SKILL.md.

Community
Advanced
cogni-daocogni-dao
4

skill-creator

Draft and refine Claude skill descriptions with evaluation loops and benchmarking.

Official
Advanced