REA
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Learn what AI writes, not just accept it. A Claude Code skill that turns AI-generated code into educational deep dives.
如果你是作者,可把这段代码贴到 README。徽章会链接回本页。
[](https://mcp.popnx.com/en/skills/mohi-devhub-antivibe)在同一工作中把 AI 连到服务的 MCP
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Codebase intelligence for TypeScript and JavaScript, exposed as structured MCP analysis tools.
MCP server for terminal commands, file operations, and process management
同类工作,排序与 Skill 列表相同
| 本项 antivibe | rea | andrej-karpathy-skills | graphify | MoneyPrinterTurbo | |
|---|---|---|---|---|---|
| 作者 | mohi-devhub | morluto | multica-ai | Graphify-Labs | harry0703 |
| Skill 数 | 1 | 1 | 1 | 1 | 1 |
| GitHub 星标 | 1,138 | 467,235 | 234,125 | 149,873 | 144,332 |
| 最近代码更新 | 2026年7月31日 | 2026年11月18日 | 2026年4月20日 | 2026年11月18日 | 2026年11月18日 |
| 许可证 | MIT | MIT | – | Apache-2.0 | MIT |
mohi-devhub 提供的 1 个 Skill(指令集),AI 读取后照做 · 不是 MCP server
MCP = 手
把 AI 连到外部服务或数据,例如读取 Notion 或开发票。需要有程序或服务器在运行。
Skill = 手册
教 AI 怎么做某项工作的文本文件,不连接任何服务。
有些会重叠:少数 Skill 带有调用工具的脚本,少数 MCP 只提供知识。请从你想完成的工作出发。
Claude Code
依据开发者 README 和 repo 中的文件夹
GitHub account mohi-devhub (person), 2 years old
★ 1,138 stars · code last changed 2026年7月31日
4 scripts that can run on a computer: scripts/analyze-code.sh, scripts/capture-phase.sh, scripts/find-resources.sh, scripts/generate-deep-dive.sh
No internet calls found in the scripts
Installing the maker’s way adds hooks that run without you asking: run every time a sub-agent finishes · run every time the AI finishes a reply and adds 2 sub-agents
Checked automatically from the repo files at commit 9991d12 (read only, nothing was run). Not a safety guarantee: read it and decide for yourself
第一步:你用哪个 AI?
选择你用的应用,本页只显示该应用的安装方法
Uploading to claude.ai would fail: the description is over 200 characters (238). Use "Let the AI install it" instead
复制这句话,粘贴到 Claude Code 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
重新打开 Claude Code,输入 /antivibe,或直接提出相关任务,Claude 会自动选用
Checked against official docs 2026年11月12日 · Source: code.claude.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/mohi-devhub/antivibe.git /tmp/popnx-skill-mohi-devhub-antivibe-9991d12 && mkdir -p ~/.claude/skills && cp -R '/tmp/popnx-skill-mohi-devhub-antivibe-9991d12/.' ~/.claude/skills/antivibeChatGPT 桌面应用内置 Codex,Codex 从 ~/.agents/skills 文件夹读取 Skill
在 ChatGPT 桌面应用中打开 Codex,把这句话粘贴到对话中
The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
把 antivibe 文件夹放到 ~/.agents/skills/antivibe
~ 是你的主文件夹($HOME)
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/mohi-devhub/antivibe.git /tmp/popnx-skill-mohi-devhub-antivibe-9991d12 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-mohi-devhub-antivibe-9991d12/.' ~/.agents/skills/antivibeChecked against official docs 2026年11月15日 · Source: learn.chatgpt.com
不能使用你自己安装的 Skill,只能使用目录中插件自带的 Skill
想在 ChatGPT 中使用这个 Skill,请按上面的方法使用 ChatGPT 桌面应用
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
复制这句话,粘贴到 Codex 的对话中 · 适用于 ChatGPT 桌面应用中的 Codex、Codex CLI 和 IDE 扩展
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Codex 会自动发现新安装的 Skill;如果没有出现,请重启 Codex · 使用方法:在 ChatGPT 中输入 @ 并选择 Skill · 在 Codex CLI 或 IDE 扩展中运行 /skills,或输入 $ 加 Skill 名称
Checked against official docs 2026年11月15日 · Source: learn.chatgpt.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/mohi-devhub/antivibe.git /tmp/popnx-skill-mohi-devhub-antivibe-9991d12 && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-mohi-devhub-antivibe-9991d12/.' ~/.agents/skills/antivibe复制这句话,粘贴到 Cursor 的对话中
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
在 Cursor 的 Agent 聊天中输入 / 并选择 antivibe
Checked against official docs 2026年11月12日 · Source: cursor.com
Skill 是 AI 会照做的指令。安装前先阅读文件,只安装可信开发者的 Skill。
Mac:按 Command+空格输入 Terminal · Windows:下方命令在 PowerShell 中无法使用,请改用上方的“让 AI 帮你安装”
这条命令把 repo 下载到临时文件夹,只复制含 SKILL.md 的文件夹
git clone --depth 1 https://github.com/mohi-devhub/antivibe.git /tmp/popnx-skill-mohi-devhub-antivibe-9991d12 && mkdir -p ~/.cursor/skills && cp -R '/tmp/popnx-skill-mohi-devhub-antivibe-9991d12/.' ~/.cursor/skills/antivibe适用于其他能读取 Skill 的应用。这句话列出了 Claude Code、Codex 和 Cursor 的路径,AI 会按你的应用选择 · Copy this sentence into the app’s chat
For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing
The AI tells you what it will do first and asks your permission at every step
Checked against official docs 2026年11月12日 · Source: code.claude.com, learn.chatgpt.com, cursor.com
其他支持 Agent Skills 的应用(如 Gemini CLI、GitHub Copilot)使用不同文件夹,请查看该应用的指南,再把含 SKILL.md 的文件夹复制过去。
已于 2026年11月11日 对照 Claude Code、Codex 和 Cursor 官方指南核实 skill 文件夹
✓ 命令已检查 2026年11月18日
以下文字来自开发者的 repo,不是 PopNx MCP 的指令
显示前 8,000 个字符 · 在 GitHub 打开完整文件
---
name: antivibe
description: Code learning and audit framework. Analyze any codebase — new, legacy, or AI-generated — and produce educational explanations or architectural audits. Use when the user wants to understand WHAT and WHY behind any code, not just accept it.
triggers:
- phrase: "/antivibe"
- phrase: "deep dive"
- phrase: "anti-vibecode"
- phrase: "why did AI write"
- phrase: "learn from this code"
- phrase: "understand what AI wrote"
- phrase: "explain what AI wrote"
- phrase: "walk me through"
- phrase: "explain this file"
- phrase: "explain this codebase"
- phrase: "analyze this module"
- phrase: "audit this"
- phrase: "just the trade-offs"
- phrase: "what should I worry about"
- phrase: "code review"
---
# AntiVibe - Code Learning & Audit Framework
## Purpose
AntiVibe generates **learning-focused explanations or architectural audits** of any code — AI-generated, legacy, or otherwise. It helps developers understand:
- **What** the code does (functionality)
- **Why** it was written this way (design decisions)
- **When** to use these patterns (context)
- **What alternatives** exist (broader knowledge)
Works on any codebase — you don't need recent git history or AI-authored files.
## When to Use
Use AntiVibe when:
1. **Manual invocation**: User types `/antivibe` or "deep dive"
2. **Post-task learning**: After a feature/phase completes, user wants to learn from it
3. **Legacy codebases**: User wants to understand existing code they didn't write
4. **Proactive**: User says "explain what AI wrote", "walk me through", "audit this", or points at a file/directory
## What AntiVibe Produces
Output saved to `deep-dive/` folder as markdown:
```
deep-dive/
├── auth-system-2026-01-15.md
├── api-layer-2026-01-15.md
└── database-models-2026-01-15.md
```
The exact sections depend on the output mode (see [Output Mode](#output-mode)):
| Section | `compact` (default) | `full` |
|---------|:---:|:---:|
| **Overview** — what the code does and why it exists | ✅ | ✅ |
| **Key Components / Concepts** — design patterns, algorithms, CS concepts used | ✅ | ✅ |
| **Code Walkthrough** — file-by-file, line-by-line notes | — | ✅ |
| **Learning Resources** — curated docs, tutorials, videos | — | ✅ |
| **Related Code** — links to other files in the codebase | — | ✅ |
## Configuration
### Known Concepts (Skip List)
Concepts listed here will not be explained in full — the explainer will only note that they were used and in what context. Edit this list to match your current knowledge.
```yaml
known_concepts:
- async/await
- React hooks
- REST APIs
```
### Output Mode
Controls how much detail is generated per run. Default is `compact` to keep token costs low.
```yaml
output_mode: compact
```
| Mode | What's included |
|------|----------------|
| `compact` (default) | Overview, key components (function-level, one line each), concepts (what + why only). No resources. No line-by-line. Max 5 files. |
| `full` | Everything in compact, plus: line-by-line walkthrough, prerequisites, curated resources, Next Steps. |
Override inline in your request:
- `"/antivibe full"`, `"full deep dive"`, `"include resources"` → `full` mode
- Default: `compact`
### Default Skill Level
Sets the explanation depth when no level is specified in the request. Options: `junior`, `mid`, `senior`. Default: `mid`.
```yaml
default_level: mid
```
| Level | Behavior |
|-------|----------|
| `junior` | Define all terms. Use analogies. Explain language features. Show full code snippets with inline comments. |
| `mid` | Skip basics. Focus on design decisions and trade-offs. Brief code references only. |
| `senior` | Skip obvious patterns. Focus only on non-obvious choices, edge cases, and architectural trade-offs. |
Level can also be specified inline in the request:
- `"explain for a junior"`, `"I'm new to this"` → `junior`
- `"I know the basics"`, `"mid level"` → `mid`
- `"senior mode"`, `"skip the basics"`, `"just the trade-offs"` → `senior`
---
## Workflow
### Step 0: Apply User Configuration
Before analyzing, read the configuration above:
- Load the `known_concepts` skip list. Any concept in this list will be acknowledged in one sentence instead of fully explained.
- Detect the skill level: check the user's request first (inline phrases take priority), then fall back to `default_level`. Apply this level consistently throughout the entire output.
- If level = `senior`, route to `agents/auditor.md` instead of continuing this workflow.
### Step 1: Identify Code to Analyze
Use the first applicable mode:
1. **Explicit** — User named specific files, a directory, or a module in their request → use those directly. No git needed. Example: "explain `src/auth/`" or "walk me through `api/routes.py`".
2. **Recent** — No explicit target given, project is a git repo, and `git diff HEAD` has output → use those changed files (current behavior for post-AI-task learning).
3. **Scan** — No explicit target, no usable git diff (legacy project, no recent changes, or not a git repo) → ask the user: "Which file, directory, or module would you like to analyze?" Do not attempt to guess.
> The code does not need to be AI-generated. AntiVibe analyzes any code.
### Step 2: Analyze Code Structure
For each file:
- Identify main purpose and responsibilities
- Note key functions, classes, modules
- Identify design patterns used (factory, singleton, observer, etc.)
- Find any complex logic or algorithms
### Step 3: Explain Concepts
For each concept/pattern found:
- **What**: Plain-language explanation
- **Why**: Why this approach was chosen over alternatives
- **When**: When to use this pattern (with context)
- **Alternatives**: Other approaches and trade-offs
- **Prerequisites**: 2–4 foundational concepts the developer must understand first (e.g., "To understand JWT, you need: HTTP request/response, Base64 encoding, cryptographic signing")
### Step 4: Find External Resources
**Only run this step in `full` mode.** Skip entirely in `compact` mode.
Search for and include:
- Official documentation for libraries/frameworks used
- Quality tutorials or blog posts
- Video resources (if available)
- Related concepts for further learning
### Step 5: Generate Output
Create markdown file in `deep-dive/` folder:
- Name format: `[component]-[timestamp].md`
- Detect output mode from the request or `output_mode` config (default: `compact`)
- **Compact mode**: Use the compact template. No line-by-line, no resources, no Next Steps. Max 5 files — if more are in scope, summarize extras in one line each and offer to go deeper.
- **Full mode**: Use the full template from `templates/deep-dive.md`. Include all sections. No 5-file limit — analyze every file in scope; for very large inputs, split the output across multiple deep-dive files rather than truncating.
- Make it educational, not just descriptive
## Auto-Trigger Configuration
AntiVibe can be configured to auto-trigger via hooks:
- **SubagentStop**: After a Task completes a feature
- **Stop**: At session end
To enable auto-trigger, configure hooks in your project (see `hooks/hooks.json`).
## Principles
1. **Why over what** - Always explain design decisions
2. **Context matters** - Explain when/why to use patterns
3. **Curated resources** - Quality links, not random Google results
4. **Phase-aware** - Group by implementation phase
5. **Learning path** - Suggest next steps for deeper study
6. **Concept mapping** - Connect code to underlying CS concepts
## Dependencies
Optional scripts in `scripts/` folder:
- `capture-phase.sh` - Detect implementation phase boundaries
- `analyze-code.sh` - Parse code structure
- `find-resources.sh` - Search for external resources
- `generate-deep-dive.sh` - Create markdown output
These are helpers - you can also do everything via direct code analysis.
## Examples
**Input**: "Explain the auth system Claude wrote" *(recent AI code)*
→ Mode: Recent (git diff). Output: `deep-dive/auth-system-2026-0提问、分享技巧或报告无法使用。按“有帮助”数量排序。