skill-creator

Create, evaluate, and iteratively improve Claude skills with benchmarked test runs.

Updated Aug 8, 2026
One-click install
npx skills add https://github.com/wolfstar-project/code-zero --skill skill-creator-wolfstar-project
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-creator
Source: https://github.com/wolfstar-project/code-zero/tree/main/.agents/skills/skill-creator
Command: npx skills add https://github.com/wolfstar-project/code-zero --skill skill-creator-wolfstar-project

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and assets (resource) components.

What problem does it solve? Writing a skill that triggers reliably and produces good outputs is hard to do by intuition alone. This Skill guides the full lifecycle of skill development: drafting SKILL.md files, running test prompts with and without the skill, grading results against assertions, and optimizing the description for accurate triggering. ## Core Features & Use Cases - Guided Skill Authoring: Interviews the user about intent, trigger contexts, and output formats, then writes a structured SKILL.md following progressive disclosure and writing-style best practices. - Evaluation & Benchmarking: Runs test prompts in parallel with-skill and baseline configurations, grades outputs against assertions, and aggregates pass rates, timing, and token usage into benchmark reports with an interactive HTML viewer. - Description Optimization: Generates realistic trigger/no-trigger eval queries, reviews them with the user, and runs an automated train/test optimization loop to improve skill triggering accuracy. - Use Case: You want a skill that formats weekly sales spreadsheets. This Skill helps you draft it, creates test prompts, runs them with and without the skill, shows you side-by-side results in a browser viewer, and iterates until the outputs meet your expectations. ## Quick Start Help me create a new skill that converts meeting notes into structured action items, then test and refine it.

Frequently Asked Questions about skill-creator

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

FAQPage Schema
How do I create a new Claude skill from scratch?▼

Start by defining what the skill should do, when it should trigger, and its output format. Then write a SKILL.md with YAML frontmatter containing a name and description, followed by markdown instructions, and optionally add scripts, references, or assets directories.

How do I test whether my skill actually improves outputs?▼

Run each test prompt twice in parallel: once with the skill and once without as a baseline. Grade both outputs against objective assertions, then aggregate pass rates, timing, and token usage into a benchmark comparing the two configurations.

How do I improve a skill description for better triggering?▼

Generate 20 realistic eval queries mixing should-trigger and should-not-trigger cases, then run the optimization loop. It splits queries into train and held-out test sets, evaluates candidate descriptions, and selects the best by test score to avoid overfitting.

What makes a good skill description for triggering?▼

A good description states both what the skill does and specific contexts when to use it, since all triggering information lives in the description. Slightly assertive phrasing helps because models tend to undertrigger skills when descriptions are too passive.

When should a skill include scripts versus inline instructions?▼

Bundle a script when test runs show subagents repeatedly writing the same helper code, or when a task is deterministic and repetitive. Scripts execute without loading into context, saving tokens and ensuring consistent behavior across invocations.

What are the limitations of skill evaluation on Claude.ai?▼

Claude.ai lacks subagents, so test prompts run sequentially without baseline comparisons, and quantitative benchmarking is skipped. Description optimization requires the claude CLI and is unavailable; evaluation relies on qualitative user feedback instead.