This skill provides a robust framework for managing and reducing Large Language Model (LLM) API costs in production environments. It introduces architectural patterns like complexity-based model routing—directing simpler tasks to cost-effective models like Claude Haiku while reserving Claude Sonnet for complex reasoning. By integrating immutable cost tracking, smart retry logic that avoids wasting tokens on permanent failures, and prompt caching for long system instructions, it allows developers to build high-performance AI agents that remain within strict budgetary constraints without sacrificing quality.
主要功能
01Prompt caching implementation for high-volume system prompts
02151 GitHub stars
03Immutable budget tracking with fail-fast guardrails for API spend
04Narrow retry logic that ignores permanent failures to save budget
05Standardized cost record structures for auditing and monitoring
06Intelligent model routing based on task complexity and length thresholds
使用场景
01Multi-model architectures needing automated selection between Haiku and Sonnet
02Production AI agents requiring strict per-user or per-session budget limits
03Batch processing large datasets where token costs accumulate rapidly