What is a Claude Code Skill?

A Claude Code Skill is a specialized capability that provides domain-specific guidance, best practices, and implementation patterns to enhance Claude's behavior during specific tasks.

How does the evaluation process work?

The tool spawns dual subagents—one with the skill and one without (the baseline)—to compare outputs directly, allowing you to measure the skill's impact both qualitatively and quantitatively.

Can I use this if I am new to terminal-based AI tools?

Yes, the Skill Developer is designed to adjust its communication style based on your familiarity, explaining technical terms like 'JSON' or 'assertions' when necessary.

What files are required to create a new skill?

At minimum, a skill requires a SKILL.md file with YAML frontmatter. This tool helps you generate that file and can also manage bundled scripts and reference documents.

How do I ensure my skill triggers at the right time?

The tool includes a dedicated 'description improver' phase that optimizes your skill's metadata to ensure Claude recognizes when the skill is relevant to your prompt.

Skill Developer & Evaluator

Name: Skill Developer & Evaluator
Author: nsalvacao

bynsalvacao

0•

Herramientas para Desarrolladores

Streamlines the creation, testing, and optimization of custom Claude Code skills through iterative evaluation and performance benchmarking.

This skill provides a comprehensive framework for building high-quality Claude Code capabilities from scratch or improving existing ones. It guides users through capturing intent, drafting precise instructions in SKILL.md, and running rigorous A/B evaluations to compare performance against baselines. By automating the setup of test cases and quantitative assertions, it ensures that new skills are reliable, trigger accurately, and deliver consistent, production-grade results through a structured lifecycle of drafting, testing, and refining.

Características Principales

01Dual-track subagent execution for A/B performance benchmarking

020 GitHub stars

03Automated drafting of SKILL.md files with optimized metadata

04Integrated quantitative assertion testing and qualitative review scripts

05Hierarchical resource management for scripts, templates, and documentation

06Skill description optimization to improve triggering accuracy and prevent 'undertriggering'

Casos de Uso

01Scaling test sets to benchmark AI performance across complex multi-step processes

02Refining and debugging existing skills that are failing to trigger or producing inconsistent results

03Building specialized workflows for domain-specific coding tasks or company standards

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nsalvacao/nsalvacao-claude-code-plugins skill-development

For use in Claude.ai and ChatGPT

Características Principales

01Dual-track subagent execution for A/B performance benchmarking

020 GitHub stars

03Automated drafting of SKILL.md files with optimized metadata

04Integrated quantitative assertion testing and qualitative review scripts

05Hierarchical resource management for scripts, templates, and documentation

06Skill description optimization to improve triggering accuracy and prevent 'undertriggering'

Casos de Uso

01Scaling test sets to benchmark AI performance across complex multi-step processes

02Refining and debugging existing skills that are failing to trigger or producing inconsistent results

03Building specialized workflows for domain-specific coding tasks or company standards

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nsalvacao/nsalvacao-claude-code-plugins skill-development

For use in Claude.ai and ChatGPT