progressive-metadata

What problem does it solve?

This Skill provides a blueprint for designing AI skills that drastically reduce token usage and operational costs. By leveraging YAML frontmatter and progressive disclosure, it ensures that only essential information is loaded initially, saving up to 84% in token costs and speeding up AI responses.

Core Features & Use Cases

• Token-Efficient Design: Structures SKILL.md files with YAML frontmatter for lightweight metadata, reducing system prompt tokens by up to 84%.
• Progressive Disclosure: Full skill instructions and supporting files (examples, templates) are loaded only when explicitly needed, optimizing context.
• Cost & Speed Optimization: Achieves 84% cost reduction and 62% faster response times compared to traditional, full-context prompting.
• Use Case: When creating a new skill for "PDF field extraction," you can define its core purpose and triggers in the frontmatter. The AI loads this small metadata first, and only if the task matches, it then loads the detailed extraction instructions and references to advanced examples, ensuring maximum efficiency.

Quick Start

Example: SKILL.md with YAML frontmatter

---

name: your-skill-name description: Comprehensive description including what the skill does, when to use it, and trigger keywords. This is loaded first and determines if the skill activates. allowed-tools: Read, Write, Grep # Optional

Your Skill Name

Full instructions here - loaded only if skill activates