skill-creator_skill

Overview

This skill defines standards for creating High-Density Agent Skills with optimal token economy. It prescribes a progressive-loading structure, strict size limits, and a compression-first writing style to keep agent contexts small and actionable. Use it to produce compact, testable skills across languages and frameworks.

How this skill works

Inspect project metadata and triggers to decide when to activate the skill. Provide a three-level content strategy: minimal metadata first, a concise core document second, and detailed examples/assets loaded on demand. Enforce size limits, imperative phrasing, and folder separation so agents load only what they need.

When to use it

• Authoring new agent skills that must be token-efficient
• Refactoring oversized documentation into lean core + on-demand references
• Preparing skills for multiple agents with different context windows
• Generating deterministic scripts, validators, or templates that run outside context
• Standardizing content layout across languages and frameworks

Best practices

• Design progressive loading: metadata → core document → references/assets
• Write imperatively: start sentences with verbs and compress wording
• Move heavy examples, long code, and tables to lazy-loaded references
• Limit core document length and inline code to strict line budgets
• Use short labels, keywords, and file-pattern triggers for proactive activation

Example use cases

• Create a compact Android skill with metadata triggers and large examples in references
• Refactor a verbose framework guide into a 40–60 line core with examples moved out
• Add a scripts/ folder for deterministic code generators and validators that execute separately
• Tune a TypeScript skill to fit 4k and 32k context windows by auditing token usage
• Standardize skill layout across React, Flutter, and Spring Boot projects