skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or iterate on skill quality. Triggers: "create a skill", "make a new skill", "build a skill for", "write a skill that", "skill for doing X", "I want a skill to", "new skill", "design a skill", "scaffold a skill", "improve this skill", "optimize
Security Assessment
About skill-creator
The skill-creator skill is a meta-skill for creating, improving, and evaluating agent skills. It solves the problem of building high-quality, reliably-triggering skills by guiding the full lifecycle: planning what a skill should do, writing SKILL.md and reference files, scoring quality against a rubric, and iterating until production standards are met. Its guiding philosophy is that a great skill is precise rather than long, with exhaustive triggers, explicit defaults, clear steps with exit gates, deferred complexity in reference files, and a structured output template.
The workflow classifies each request into Create, Improve, or Evaluate mode. Create mode gathers requirements (task, target user, tools/APIs, inputs, outputs, credentials, platform), then plans an architecture by choosing a structural pattern (Linear, Router, Methodology, Widget, or API Wrapper) and a five-to-nine step outline. A core rule is 'always dynamic, never static': skills must detect available tools, libraries, and auth at runtime and adapt via decision trees and fallback paths rather than hardcoding one method. Reference files document dynamic-calling patterns, architecture patterns, a complete YAML frontmatter guide (including environment variables, credential files, and conditional activation), a ten-dimension quality rubric, annotated skill examples, and a writing guide. It encodes patterns distilled from analyzing many production finance and agent skills.
It targets skill authors and prompt/agent engineers who want to scaffold new skills, refactor weak ones, or benchmark skill quality. It runs on Claude Code and other CLI agents (file creation requires a CLI) and on Claude.ai for evaluation and planning. It is a documentation and authoring aid with no risky operations.
FAQ
What modes does it support?
Three: Create a brand-new skill, Improve/fix an existing skill, and Evaluate/score a skill against a ten-dimension quality rubric. If the request is ambiguous it asks which you want.
What is the 'always dynamic, never static' rule?
Skills must detect what tools, libraries, and auth are available at runtime and adapt with decision trees and fallback paths, rather than hardcoding a single method. The dynamic-calling reference documents the full pattern catalog.
Where does it run?
On Claude Code and other CLI-based agents for full functionality (skill file creation requires a CLI), and on Claude.ai for evaluation and planning.
What structural patterns can a new skill use?
Linear, Router, Methodology, Widget, and API Wrapper, each with guidance on when to use it and a target of five to nine steps before splitting into a router pattern.
Install skill-creator
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
himself65/finance-skills