wtf is claude-code-skills?

This repository is a collection of skills for Claude Code, the idea being reusable, documented workflows that make AI assisted coding sessions cheaper or faster to run. Right now it contains one skill, called multi-tier-models, which is about using different AI models for different parts of a task instead of relying on a single model throughout. The approach behind multi-tier-models is to assign the biggest, most capable models to high level reasoning and planning, and hand off the more repetitive execution work, like writing text or code, to smaller and cheaper models, including one described as a multimodal model used for analyzing images. According to the numbers given in the README, this combination cut costs by 83 percent on one test, roughly from 1.89 dollars down to 0.32 dollars, reduced the time taken by 75 percent, from about 8 minutes to 2 minutes, and used 64 percent fewer tokens, while still producing output of comparable quality to using one large model for everything. To use it, you clone the repository and copy the multi-tier-models folder into your local Claude Code skills directory, after which it becomes available as a slash command inside Claude Code. The README also describes a manual setup option, where you create the skill folder yourself and copy the files in directly, rather than cloning the whole repository. The project currently lists just this one skill in a summary table, along with its purpose and reported savings, and notes that it is active and ready to use. It is maintained by one author and is released under the MIT License, with the skill's own documentation available in a separate file inside its folder for anyone who wants the full details of how the model tiers are chosen and applied.