skill-builder

Create Claude Code skills with scaffolded files and evaluated triggers.

Updated May 17, 2022
One-click install
npx skills add https://github.com/kanade0404/dotfiles --skill skill-builder-kanade0404
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-builder
Source: https://github.com/kanade0404/dotfiles/tree/main/.opencode/skills/skill-builder
Command: npx skills add https://github.com/kanade0404/dotfiles --skill skill-builder-kanade0404

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill helps you create, diagnose, and improve Claude Code skills without guessing why they do or do not fire, so trigger behavior and post-activation quality can be tuned in one place.

Core Features & Use Cases

  • Scaffolds new skills in either consumer project directories or rulesync source catalogs.
  • Designs should-use and should-skip eval cases, then measures trigger precision and recall.
  • Refines descriptions to reduce false positives and false negatives while keeping scope project-agnostic.
  • Reviews activated skill bodies with a lightweight quality loop and records recurring failure patterns.
  • Use it when a team needs a new skill, a missed trigger fixed, or an overbroad skill narrowed.

Quick Start

Ask this skill to scaffold a new skill or tune an existing one, then provide the target task, current skill path, and the main misfire examples.

Frequently Asked Questions about skill-builder

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

FAQPage Schema
How do I fix Claude Code skills that fail to trigger or fire too often?

Fix Claude Code skills by designing should-use and should-skip eval cases to measure trigger precision and recall, then refine descriptions to reduce false positives and false negatives.

How do I scaffold new skills for Claude Code projects?

Scaffold new skills by generating skill files in either consumer project directories or rulesync source catalogs, then apply frontmatter validation to ensure correct structure.

Why does my Claude Code skill description cause false positive activations?

Skill descriptions trigger false positives when scope is too broad; tune trigger behavior by iterating on descriptions with eval cases to keep scope project-agnostic.

Can I use this to review the quality of an activated skill body?

Review activated skill bodies using a lightweight quality loop that identifies recurring failure patterns and assesses post-activation quality in local skill directories.

What's the best way to evaluate Claude Code trigger scoring for skills?

Evaluate Claude Code trigger scoring by defining should-use and should-skip cases, measuring precision and recall, and iterating on descriptions to improve accuracy.

Do I need a specific framework to tune prompt engineering evals for skills?

No specific framework is needed; this project-agnostic tuning process requires only local skill directories to run lightweight eval loops and frontmatter validation.