What is the 'info tool pattern' in MCP?

The info tool pattern provides a single entry point (typically an 'info' tool) that allows users to explore capabilities progressively via categories and summaries rather than receiving a massive dump of all available tools at once.

How does the skill handle AI hallucinations?

It implements an 'Accept Extra Parameters' pattern that warns about unknown parameters instead of rejecting them, making the tool more resilient to common AI agent behaviors.

What are the four levels of progressive detail?

The levels include Level 1 (Counts only), Level 2 (Names and one-line descriptions), Level 3 (Detailed input/output schemas), and Level 4 (Full documentation with examples).

How does this skill help save tokens in Claude Code?

By implementing sparse overviews that use as few as 20-50 tokens, the AI only loads the full, verbose documentation for specific tools when requested, preventing context window bloat.

When should I use layered discovery instead of a simple tool list?

Layered discovery is recommended whenever an MCP server exceeds 15 tools or when tools are grouped into distinct functional categories that require contextual guidance.

Progressive Discovery

Name: Progressive Discovery
Author: standardbeagle

bystandardbeagle

•

API開発

Implements a layered information architecture for MCP servers to help users navigate complex toolsets through intuitive discovery patterns.

概要

The Progressive Discovery skill provides a standardized framework for implementing 'info' and 'help' tools within MCP servers, solving the problem of tool discovery in complex environments. By utilizing a tiered approach—moving from sparse overviews to detailed documentation—it prevents token waste and cognitive overload for AI agents and human users alike. This pattern includes templates for sparse tables, machine-readable JSON outputs, and common workflow guides, ensuring that users can find the exact capability they need without sifting through exhaustive documentation for every single interaction.

主な機能

Multi-layered discovery patterns ranging from sparse overviews to full tool documentation
Token-optimized output formats including sparse tables and machine-parseable JSON
Pre-defined workflow templates to guide users through common multi-tool sequences
Scalable implementation guides for MCP servers ranging from 5 to 20+ tools
Permissive parameter handling patterns to mitigate AI model hallucinations
1 GitHub stars

ユースケース

Managing complex MCP servers with high tool counts that overwhelm standard documentation
Onboarding users to new MCP capabilities through structured discovery and help systems
Reducing token consumption in AI-driven workflows by providing layered info access

概要

主な機能

Multi-layered discovery patterns ranging from sparse overviews to full tool documentation
Token-optimized output formats including sparse tables and machine-parseable JSON
Pre-defined workflow templates to guide users through common multi-tool sequences
Scalable implementation guides for MCP servers ranging from 5 to 20+ tools
Permissive parameter handling patterns to mitigate AI model hallucinations
1 GitHub stars

ユースケース

Managing complex MCP servers with high tool counts that overwhelm standard documentation
Onboarding users to new MCP capabilities through structured discovery and help systems
Reducing token consumption in AI-driven workflows by providing layered info access