What is ai-cost-cutter-skills?

This project is a set of 10 downloadable skills that plug into AI coding assistants like Claude Code, Codex, or Cursor to help reduce how much you spend on AI usage. Instead of giving generic advice like switching to a cheaper model, each skill turns a specific cost saving idea into a concrete procedure the assistant follows, and each one includes a proof step that this repository's automated tests actually run on every code change, so the claims are checked rather than just stated. The 10 skills cover different angles on cost control. Some focus on routing, such as defaulting to a cheap model and only escalating to an expensive one for genuinely hard tasks, or capping how often a costly reviewer model gets called. Others focus on visibility, like tracking spending by both token count and dollar cost since those can tell different stories. There are skills for trimming how much text gets fed back into an assistant's context window repeatedly, for comparing models against your own real prompts before committing to one, and for planning a large batch job so it fits within a free tier's rate limits and budget. Installing the skills can be done a few ways: a single command line tool that detects which coding assistant you use and installs everything into the right place, a Claude Code plugin that installs all ten at once, or manually copying individual skill folders into your project. There is also a simpler option of adding one summary file to a project so an assistant follows the ideas as general guidance instead of specific commands. The project is honest about what its automated checks can and cannot prove. The repository's own tests confirm that the deterministic parts of each skill, like configuration and math, actually work as described, but it states clearly that real world quality and cost differences on someone's specific work must still be measured by that person, not assumed from the badge. This project targets developers already using AI coding assistants who want a more evidence based way to manage what those assistants cost them. It is released under the MIT License.