Skill UI

This guide details advanced strategies for optimizing Claude API usage to reduce latency and cost. Techniques covered include prompt caching (significant cost reduction), strategic model selection (Haiku, Sonnet, Opus) based on task complexity, implementing streaming for perceived speed, and managing token limits. Use this guide when building high-performance, cost-efficient LLM applications that interact with large language models.

A comprehensive guide to minimizing expenses when utilizing the Anthropic Claude API. Learn advanced strategies including right-sizing models (Haiku, Sonnet, Opus), prompt caching (up to 90% savings), message batching, and effective token reduction techniques. Essential for building cost-efficient AI applications.

This skill activates 'Steno Mode,' providing highly compressed, expert-level responses by eliminating filler words and superfluous tokens (up to 40% reduction). It maintains absolute technical precision, making it ideal for technical communications like code reviews, debugging Q&A, and architecture summaries. Users can select four defined levels (lite, brief, court, machine) to control the density, with exactness always taking top priority.

Consent-gated editor for Caveman Learn. Reads ranked token sinks, proposes concrete reductions for heavy context files and repeated blocks, applies them after approval. Use to lower agent token costs, trim heavy CLAUDE.md, offload re-pasted context to cavemem, or distill repeated procedures into a skill.