XiNian-dada/Fuck_My_Shit_Mountain

Overview

Fuck My Shit Mountain is a community‑contributed skill for AI‑powered coding assistants such as Codex, Claude Code, GitHub Copilot, and Gemini. The skill lets the assistant act as a systematic code‑audit “agent”: it scans a repository, builds a lightweight inventory of the stack and dependencies, then runs a checklist of 26+ engineering dimensions (security, concurrency, performance, privacy, AI‑safety, etc.). For each finding it supplies:

• the exact file and line number,
• the concrete trigger condition,
• an impact rationale and an effort estimate,
• a prioritized remediation plan with a minimal‑fix suggestion and a regression‑test tip.

The output can be rendered as an interactive HTML page, a JSON blob for CI pipelines, or a simple Markdown summary.

---

Key Features

Feature	What it does
Automated project portrait	Detects language, framework, dependencies, and high‑risk surfaces (LLM usage, DB access, APIs) and suggests relevant audit dimensions.
Evidence‑based findings	Every issue is tied to a concrete code location and a reproducible trigger, never a vague “add a comment”.
Risk categorisation	Separates confirmed defects from “to‑be‑validated” items; untouched areas are marked Not assessed.
Incremental mode	Works on a git diff (e.g., PR changes) so it can be plugged into daily code‑review workflows.
Multi‑format reports	One‑click export to interactive HTML, machine‑readable JSON, or lightweight Markdown.
26+ audit dimensions	Includes security, stability, concurrency, performance, testing authenticity, maintainability, AI‑safety, supply‑chain, cost, privacy, and many more.
Agent‑agnostic installation	Simple copy‑paste into the skill directory of Codex, Claude, Copilot, or Gemini; also supports a “send the repo link to the AI IDE” shortcut.

---

How to Use

1. Install the skill – either ask your AI IDE to fetch the repo (https://github.com/XiNian-dada/Fuck_My_Shit_Mountain) or copy the fuck‑my‑shit‑mountain/ folder into the appropriate ~/.<agent>/skills/ path.
2. Trigger an audit with natural language, e.g.:

   请使用 fuck‑my‑shit‑mountain 对当前项目进行全量审计。
   报告语言：中文
   输出格式：html
   

   The assistant maps the request to an internal mode (full, security, incremental, etc.) and runs the analysis.
3. Review the report – the generated HTML shows a score panel (Security, Stability, Performance, …) and a list of top risks with line‑level evidence and remediation steps.
4. Integrate into CI – consume the JSON output to fail a build if any dimension falls below a threshold, or to automatically create a ticket with the suggested fix.

---

Who Might Benefit?

• Engineering managers who need a fast, repeatable “first pass” audit before a release.
• Security or reliability teams looking for a structured checklist that can be run by an LLM‑assistant.
• Developers using AI pair‑programmers who want the assistant to go beyond style suggestions and surface hidden architectural or safety bugs.

---

Limitations

• The skill is read‑only; it never modifies source files or commits changes.
• It does not replace a deep manual security review; it surfaces likely issues and prioritises them for human follow‑up.
• Effectiveness depends on the underlying LLM’s ability to understand the codebase; very large repositories may need the incremental mode to stay within context limits.

---

License

MIT License – free to use, modify, and redistribute.

---

All details are taken directly from the repository’s README; no additional features have been inferred.