What is fuck_my_shit_mountain?

This repository is a skill package for AI coding assistants, designed to run structured code audits and produce detailed reports. The README is written primarily in Chinese. The core idea is a prompt-driven framework that you install into a supported AI tool, then invoke to analyze a codebase and get back a scored report organized by category. The framework supports 15 named audit modes that can be used individually or combined. These include modes focused on security, performance, testing quality, code maintainability, type safety, frontend state management, and backend API design, among others. A full mode runs all 19 dimensions at once. Each audit produces scores on a 0 to 10 scale per dimension, where 10 means clean code and lower scores indicate increasing problems. The output format can be plain Markdown, an HTML page with sidebar navigation and color-coded score bars, or both. The HTML demo page included in the repository is labeled clearly as a fictional example rather than a real audit result. The skill is structured as a directory containing a main entry file, 15 mode-specific prompt files, scoring rubrics defining severity levels and evidence standards, and report templates. Installation involves copying this directory into the skills folder of whichever AI tool you use: the README gives instructions for Claude Code, GitHub Copilot, Codex, and Gemini CLI. After loading, you invoke it by name and specify which mode, language, and output format you want. The project description says the audits are evidence-based and structured, and the disclaimer at the top of the README notes that results are for reference only and should always be combined with human review and real-world testing. The project is released under the MIT license.