This skill provides a comprehensive framework for creating high-performance tools that AI agents can interact with reliably and efficiently. It emphasizes that tool descriptions are more critical than implementation for LLM accuracy, offering deep guidance on JSON Schema design, parameter documentation, and the Model Context Protocol (MCP). By implementing strict validation gates and informative error recovery, users can prevent agent hallucinations, reduce token costs, and build production-grade agentic workflows that scale across multiple platforms and providers.
주요 기능
01Best practices for parameter descriptions and enum usage
02Model Context Protocol (MCP) implementation patterns
0340,583 GitHub stars
04Comprehensive tool error handling and recovery strategies
05JSON Schema design optimization for LLM readability
06Validation gate design to prevent silent agent failures
사용 사례
01Refactoring existing tool definitions to reduce agent hallucination rates
02Building custom MCP servers to connect Claude to internal databases
03Developing multi-agent systems with shared tool schemas and validation