0xiehnnkta/nemesis-auditor 是从 https://github.com/0xiehnnkta/nemesis-auditor. 导入的特工技能候选人
概览
0xiehnnkta/nemesis-auditor 是从 https://github.com/0xiehnnkta/nemesis-auditor. 导入的特工技能候选人
概述（AI 翻译）
无法逃避的审计员
一个针对 Claude Code 的迭代深度逻辑安全审计代理。Nemesis 结合了两种互补的审计方法，形成交替反馈循环，以发现双方单独无法发现的漏洞。语言无关——适用于 Solidity、Move、Rust、Go、C++、Python、TypeScript 或任何代码库。
工作原理
Nemesis 是一个agent——一个在两个子代理之间进行迭代来回循环的编排器：
| 通过 | 特工 | 它发现了什么 |
|---|---|---|
| 1 | 费曼审计师 | 通过第一原则质疑出现商业逻辑漏洞。每一行都被质疑。每一个假设都被揭露。 |
| 2 | 状态不一致审计员 | 耦合状态不同步的错误。映射每一条变异路径。找出一方更新而另一方更新的空白。 |
| 3+ | 交替进行针对性通关 | 每次通关都会将发现信息传递到下一次。费曼嫌疑人成为国家审计目标。州级缺口成为费曼审讯点。 |
循环一直持续到收敛——每次扫描都没有新发现（最多6次）。
为什么是两者？
- 费曼单独发现逻辑漏洞，但可能遗漏结构状态缺口
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
N E M E S I S
The Inescapable Auditor
+===============================================================+
| |
| "Nemesis -- the goddess of divine retribution against |
| those who succumb to hubris." |
| |
| Your code was written with confidence. |
| Nemesis questions that confidence. |
| Then maps what your confidence forgot to protect. |
| Then questions it again. |
| |
| Nothing survives both passes. |
| |
+===============================================================+
An iterative deep-logic security audit agent for Claude Code. Nemesis combines two complementary audit methodologies in an alternating feedback loop to find bugs that neither can catch alone. Language-agnostic – works on Solidity, Move, Rust, Go, C++, Python, TypeScript, or any codebase.
How It Works
Nemesis is an agent – an orchestrator that runs an iterative back-and-forth loop between two sub-agents:
| Pass | Agent | What It Finds |
|---|---|---|
| 1 | Feynman Auditor | Business logic bugs via first-principles questioning. Every line challenged. Every assumption exposed. |
| 2 | State Inconsistency Auditor | Coupled state desync bugs. Maps every mutation path. Finds gaps where one side updates without the other. |
| 3+ | Alternating targeted passes | Each pass feeds findings into the next. Feynman suspects become state audit targets. State gaps become Feynman interrogation points. |
The loop runs until convergence – no new findings in a pass (max 6 passes).
Why Both?
- Feynman alone finds logic bugs but may miss structural state gaps
- State Inconsistency alone finds desync bugs but may miss WHY state was designed that way
- Nemesis runs them back and forth – each pass feeds the next – finding bugs at every iteration that the previous pass missed
Requirements
- Claude Code CLI installed and configured
- A codebase you want to audit
Installation
Option A: Copy into your project (recommended)
# Clone this repo
git clone https://github.com/0xiehnnkta/nemesis-auditor.git
# Copy the .claude folder into your project
cp -r nemesis-auditor/.claude /path/to/your-project/
# You can delete the cloned repo after copying
rm -rf nemesis-auditor
Option B: Run from this repo
# Clone and enter the repo
git clone https://github.com/0xiehnnkta/nemesis-auditor.git
cd nemesis-auditor推荐工具
换一个关键词，或者移除筛选条件。
安装
npx skillfish add 0xiehnnkta/nemesis-auditor