skills.rest
dredd-us
dredd-usOFFICIAL
Explore all skills by this author →
GitHub

progressive-metadata

Design AI skills with YAML frontmatter for progressive disclosure.

Updated Aug 23, 2026
Search Tags:#prompt engineering#skill design#AI efficiency#progressive disclosure#token optimization#cost savings
One-click install
npx skills add https://github.com/dredd-us/seashells --skill progressive-metadata-dredd-us
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: progressive-metadata
Source: https://github.com/dredd-us/seashells/tree/main/.claude/skills/progressive-metadata
Command: npx skills add https://github.com/dredd-us/seashells --skill progressive-metadata-dredd-us

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes examples (resource) components.

What problem does it solve?

This Skill teaches and demonstrates how to design AI skills using YAML frontmatter for progressive disclosure, drastically reducing token usage (84% reduction) and cost while maintaining full functionality and consistent outputs.

Core Features & Use Cases

  • Metadata-First Loading: Only lightweight metadata is loaded initially, full instructions on demand.
  • Token & Cost Savings: Reduces prompt tokens from ~8,750 to ~1,400, saving 84% per call.
  • Use Case: When creating a new skill for "PDF field extraction," use this pattern to ensure its description is rich enough for activation, but its detailed instructions are only loaded when needed, optimizing every interaction.

Quick Start

Use the progressive-metadata skill to design a new skill for "data transformation" that optimizes token usage.

Frequently Asked Questions about progressive-metadata

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

FAQPage Schema
How much can progressive disclosure reduce token usage in AI skills?▼

Progressive disclosure with YAML frontmatter reduces prompt tokens by 84%, cutting typical skill calls from ~8,750 tokens to ~1,400 tokens by loading only metadata initially and delivering full instructions on demand.

When should I use metadata-first loading for skill design?▼

Use metadata-first loading when designing skills for PDF extraction, data transformation, or protocol workflows where rich descriptions are needed for activation but detailed instructions should load only when the skill is invoked, optimizing token efficiency.

How do I design a skill that saves tokens while maintaining full functionality?▼

Structure your skill with YAML frontmatter to separate lightweight metadata from complete instructions, ensuring consistent outputs and full capability while drastically reducing tokens per interaction through progressive content delivery.

What's the difference between loading all instructions upfront versus progressive disclosure?▼

Upfront loading includes all instructions in every call, wasting tokens on unused content. Progressive disclosure loads only metadata initially, delivering detailed instructions on demand, reducing cost and latency without sacrificing functionality or output consistency.

Can I apply progressive metadata to existing skills?▼

Yes. Refactor existing skills by extracting core metadata into YAML frontmatter and converting full instructions to reference-based supporting files, enabling on-demand context loading and achieving 84% token reduction across PDF extraction, data transformation, and protocol workflows.

Related Skills

Explore More Software Engineering Skills→
ntluong95ntluong95

Agent Skills Architecture

Establishes a token-optimized CLI-driven framework for modular AI agent instructions with progressive disclosure.

Community
Advanced
pradyumnasagarpradyumnasagar
6

agent-handoff

Automate inter-skill data transfer and provenance capture via JSON Material Passports.

Community
Advanced
cablatecablate
39

agentic-mcp

Manage MCP servers via progressive disclosure and a socket daemon.

Community
Intermediate
HouseGarofaloHouseGarofalo
3

context-engineering

Optimize AI context windows and manage token budgets with RAG patterns.

Official
Advanced