skill-creator_skill

Overview

This skill guides creating and optimizing modular skills that extend Claude with domain-specific workflows, tool integrations, scripts, references, and assets. It focuses on practical steps: capturing concrete examples, planning reusable resources, initializing a skill scaffold, iterating on reusable contents, and packaging for distribution. The guidance emphasizes token-efficient design and progressive disclosure to keep runtime context small.

How this skill works

Inspect use cases and sample interactions to define the skill's scope and triggers. Recommend a small manifest and concise instructional file for quick discovery, then add modular resources—scripts for deterministic tasks, reference docs for domain knowledge, and assets for output artifacts—loaded only when needed. Validate resources with tests, ensure environment variable precedence, and package the skill after automated validation.

When to use it

• Create a new skill to automate repeated development tasks or workflows.
• Refactor an existing skill to improve token efficiency or enable progressive disclosure.
• Add deterministic scripts for tasks that are frequently rewritten (e.g., file transforms, API callers).
• Include references for schemas, policies, or API docs that Claude should consult on demand.
• Bundle assets when the skill must produce files or templates in outputs.

Best practices

• Start from concrete examples to define supported queries and triggers.
• Design a short manifest and a concise instruction file for fast automatic activation.
• Split large documentation into small reference files; load them only when required.
• Prefer Python or Node scripts over shell for cross-platform reliability and include tests.
• Respect environment variable precedence and include an .env.example to document required keys.
• Keep the instruction file and reference files token-efficient; use progressive disclosure.

Example use cases

• Create a PDF-processing skill with a reusable Python rotate-and-merge script and brief usage notes.
• Build a cloud-deploy skill that includes API reference snippets and a deploy script honoring env precedence.
• Package a data-analysis skill with table schemas in references and query scripts for repeated metrics.
• Develop a frontend-generator skill bundling a boilerplate asset folder and a short guide to customize it.