How does this skill help with token costs?

By optimizing schemas and providing crystal-clear instructions, it reduces the need for the LLM to retry failed calls or generate unnecessary reasoning steps, significantly lowering token consumption.

What anti-patterns does this skill address?

It helps developers avoid 'silent failures' where a tool fails without informing the agent, and 'tool bloat' where too many available tools confuse the model's decision-making process.

Why are tool descriptions more important than implementation?

LLMs cannot see your backend code; they only see the schema and descriptions you provide. A vague description causes the model to guess usage, leading to hallucinations and errors.

Does this skill support the Model Context Protocol (MCP)?

Yes, it incorporates emerging MCP standards, which act as a universal interface for connecting AI agents to diverse data sources, tools, and environments.

Agent Tool Builder

Name: Agent Tool Builder
Author: claudiodearaujo

byclaudiodearaujo

API開発

Optimizes tool definitions and JSON schemas to ensure AI agents interact reliably with external systems without hallucination or silent failures.

概要

Agent Tool Builder provides expert guidance for designing the critical interface between LLMs and the physical world. It focuses on the philosophy that tool descriptions and schemas are more important than the underlying code, offering best practices for JSON Schema design, the Model Context Protocol (MCP), and proactive error handling. By implementing these patterns, developers can prevent common agent failures such as infinite loops and hallucinated parameters, resulting in more robust, token-efficient, and capable AI integrations.

主な機能

Implementation of the Model Context Protocol (MCP) standard
Recovery-oriented error handling to guide agents through failures
Input examples and validation patterns to reduce model hallucination
Strategic description writing optimized for LLM comprehension
High-precision JSON Schema design for unambiguous tool definitions
0 GitHub stars

ユースケース

Refining existing function-calling schemas to reduce agent failure rates
Developing custom MCP tools for integrating local databases with Claude
Designing complex multi-tool workflows for autonomous AI agents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add claudiodearaujo/izacenter agent-tool-builder

For use in Claude.ai and ChatGPT

Download Skill

GitHub

概要