Optimizes Claude Code session costs and performance by intelligently selecting the most appropriate model for specific development tasks.
This skill provides a comprehensive framework for managing Claude Code sessions by matching task complexity with the optimal model—Opus, Sonnet, or Haiku. It features a detailed decision matrix for architecture, debugging, and routine tasks, alongside budget planning tools and cascading strategies to maximize reasoning capability while minimizing token expenditure. By implementing smart upgrade and downgrade triggers and leveraging specialized plan modes, it ensures high-stakes work gets the intelligence it needs without overpaying for mechanical or repetitive coding tasks.
Key Features
0110 GitHub stars
02Cost-per-turn budget estimation and projection
03Intelligent upgrade/downgrade logic triggers
04Multi-model cascading for phased development
05Optimized /plan mode for architectural scoping
06Task-based model decision matrix
Use Cases
01Automating bulk file reformatting using the cost-effective Haiku model.
02Deciding when to switch from Sonnet to Opus for deep root-cause debugging.
03Estimating the total budget for a complex multi-file feature implementation.