Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Dsl Vm Reverse

ASecurity

Reverse JavaScript-based custom DSL/VM interpreters, non-standard WASM-like runtimes, and risk-control engines. Use when analyzing IIFE or switch-based opcode dispatchers, extracting instruction tables, recovering bytecode semantics, capturing VM state at runtime, or reconstructing execution flow.

434 stars
0 votes
0 copies
0 views
Added 9/23/2026
ai-agentsjavascriptpythongojavabashawsapi

Works with

cliapi

Security Analysis

A100/100

Scanned 9/23/2026

$npx -y skills add coco-research/coco --skill dsl-vm-reverse --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Dsl Vm Reverse?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Dsl Vm Reverse
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/coco-research-dsl-vm-reverse/badge)](https://www.skillsdirectory.com/skills/coco-research-dsl-vm-reverse)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: dsl-vm-reverse
description: Reverse JavaScript-based custom DSL/VM interpreters, non-standard WASM-like runtimes, and risk-control engines. Use when analyzing IIFE or switch-based opcode dispatchers, extracting instruction tables, recovering bytecode semantics, capturing VM state at runtime, or reconstructing execution flow.
---

# 🔄 DSL 自定义虚拟机逆向(DSL VM Reverse Engineering)

## 目录

- [1. 适用范围](#1-适用范围)
- [2. DSL VM 识别特征](#2-dsl-vm-识别特征)
- [3. 通用逆向工作流](#3-通用逆向工作流)
- [4. Opcode 提取与分类](#4-opcode-提取与分类)
- [5. 运行时捕获方案](#5-运行时捕获方案)
- [6. 常见状态码](#6-常见状态码)
- [7. Skill 自检清单](#7-skill-自检清单)

---

## 1. 适用范围

当目标文件符合以下 **任意特征** 时使用本 skill:

| # | 特征 | 说明 |
|---|------|------|
| 1 | IIFE 开头 + 大量单字母变量名 | `!function(){var U=void 0,y=parseInt,E0=Function,...}` |
| 2 | 包含 `DG()` 或类似函数含 switch-case 循环 | 解释器主循环,`d[7]&31` 解码 opcode |
| 3 | 大文件(500KB+)但零字节占比 < 1% | 非标准 WASM,纯 JS |
| 4 | 包含 `C[number]` 常量表引用 | `C[9][xxx]` 函数表/字符串表 |
| 5 | 单行压缩代码 | 583KB 单行,混淆变量名 |

### 排除规则

| 条件 | 非本 skill | 转至 |
|------|-----------|------|
| 文件以 `\x00asm` 开头 | 标准 WASM 二进制 | `reverse-engineering/languages.md` |
| 文件以 `Uint8Array([0,97,115,109])` 含 WASM 魔术字 | WASM 嵌入式 | 提取 .wasm 后转 IDA/Ghidra |
| 标准 Webpack 打包(`function(e,t,n){...}`) | 普通 JS | `js-reverse/` |
| 零字节占比 > 20% | WASM 二进制 | `reverse-engineering/languages.md` |

---

## 2. DSL VM 识别特征

### 代码特征

```javascript
// 特征 1: IIFE 入口,单字母变量映射数字常量
!function(){
    var U=void 0, y=parseInt, E0=Function, AN=Uint8Array;
    var E=15, l=10, m=12, x=16, S=13, $=11;
    // 数字常量映射为变量名,替代原始数字
    ...
}

// 特征 2: 解释器主循环 DG()
function DG(C, d, ...) {
    var d = [];  // 数组模拟 WASM stack/locals
    for (d[7] = x; d[7] !== U;) {
        var aE = d[7] & 31;         // 低 5 位 = opcode
        var O = d[7] >> 5 & 31;      // 高 5 位 = sub-operation
        switch (aE) {
            case 0: /* ... */ d[7] = 612; break;
            case 1: /* ... */
            // ... N 个 case
        }
    }
}

// 特征 3: 常量表 C[9] 存储函数索引和字符串
// C[9][0] = ["pc"]      → 函数参数描述
// C[9][667] = "string"  → 字符串常量
// C[9][x] = number      → 函数索引

// 特征 4: W(C[index], null, ...) 调用模式
// W = Function.prototype.call.bind(call)
// 所有内置函数通过 C[index] 索引调用

// 特征 5: 指令编码格式
// d[7] = opcode(bit 0-4) | subop(bit 5-9) | operand(bit 10+)
```

### Opcode 编码格式

每条指令编码为 32 位整数:

```
bit 0-4:   opcode (0-N)
bit 5-9:   sub-operation (0-31)
bit 10-31: operand/立即数

解码:
  aE = d[7] & 31        → opcode
  O  = d[7] >> 5 & 31   → sub-operation
  d[other] = d[7] >> 10  → operand
```

---

## 3. 通用逆向工作流

### Phase 1: 文件分类(5 分钟)

```bash
# 检查是否为 DSL VM
python3 << 'EOF'
with open('target.js', 'rb') as f:
    head = f.read(100)

# 1. 检查 WASM 魔术字
if head[:4] == b'\x00asm':
    print("标准 WASM 二进制")
    exit()

# 2. 检查零字节占比
data = open('target.js', 'rb').read()
zero_pct = data.count(b'\x00') / len(data) * 100
print(f"零字节占比: {zero_pct:.1f}%")

if zero_pct > 20:
    print("WASM 二进制")
elif head[:2] == b'!f':
    # 检查单字母变量模式
    if b'var U=void 0' in head or b'U=void 0,y=parseInt' in head:
        print("→ DSL VM!")
    else:
        print("普通 JS IIFE")
EOF
```

### Phase 2: 变量映射表提取(10 分钟)

```python
import re

with open('target.js', 'r', errors='replace') as f:
    s = f.read()

# 提取开头 2000 字符的 var X=数字 映射
mappings = re.findall(r'var\s+(\w+)\s*=\s*(\d+)', s[:2000])
print('常量映射:')
for name, val in mappings:
    print(f"  {name:4s} = {val:3d} (0x{int(val):02x})")
```

### Phase 3: Opcode 提取与分类(15 分钟)

```python
# 1. 提取所有 case
all_cases = re.findall(r'case\s+(\d+):', s)
unique = sorted(set(int(c) for c in all_cases))

print(f"总 case: {len(all_cases)} 个")
print(f"唯一 opcode: {len(unique)} 个: {unique}")

# 2. 分类每个 opcode
for op in unique:
    idx = s.find(f'case {op}:')
    snippet = s[idx:idx+200]
    if 'd[7]=' in snippet:
        op_type = 'BRANCH'
    elif 'return' in snippet:
        op_type = 'RETURN'
    elif 'W(C[' in snippet:
        op_type = 'CALL'
    elif 'new' in snippet:
        op_type = 'ALLOC'
    elif 'try' in snippet or 'catch' in snippet:
        op_type = 'EXCEPTION'
    else:
        op_type = 'ARITH/STORE'
    print(f"  opcode {op:2d}: {op_type}")
```

### Phase 4: 常量表分析(30 分钟)

```python
const_refs = re.findall(r'C\[9\]\[(\d+)\]', s)
unique_refs = sorted(set(int(x) for x in const_refs))

print(f"C[9] 引用: {len(unique_refs)} 个索引")
print(f"范围: {min(unique_refs)} - {max(unique_refs)}")

# 对每个引用分析上下文
for ref in unique_refs[:20]:
    idx = s.find(f'C[9][{ref}]')
    ctx = s[max(0,idx-50):idx+80]
    clean = ''.join(c if c.isprintable() else ' ' for c in ctx)
    print(f"  C[9][{ref}] → {clean}")
```

### Phase 5: 导出函数追踪(1-2 小时)

导出函数(如 `getToken`)通过以下路径定位:

```
1. 找 AWSCInner.register() 或类似注册调用
2. 确定注册的模块和工厂函数
3. 找工厂函数返回的对象 → 导出函数定义位置
4. 若函数名不在 JS 中 → 在 C[9] 常量表中作字节码存储
5. 追踪调用链:
   AWSCInner._modules['fy'].getToken()
   → W(C[函数索引], null, ...)
   → DG() 解释器执行编码后的指令序列
```

### Phase 6: 运行时注入(若纯静态分析不够)

```javascript
// 注入最小 AWSC 兼容环境
const fakeEnv = {
    AWSCInner: {
        _modules: {},
        register(name, moduleName, factory) {
            this._modules[moduleName] = factory();
        }
    }
};

// 执行 DSL VM 代码
dslVmCode();

// 获取导出
const token = fakeEnv.AWSCInner._modules['fy'].getToken({});
```

---

## 4. Opcode 提取与分类

### 参考 opcode 对照表(基于已有案例)

| Opcode | 操作类型 | 特征 |
|--------|---------|------|
| 0 | **BRANCH** | `d[7]=xxx` 无条件跳转 |
| 1 | **CALL** | `W(C[Y],null,function(){...})` 嵌入函数调用 |
| 2 | **ARITH** | `d[4]=0`, `d[7]=72` 变量赋值 |
| 3 | **ARITH** | `d[0]=d[1][C[x]]`, `d[5]=d[0]<d[3]` 比较运算 |
| 4 | **STORE** | `d[8]=d[5]in d[4]` 属性访问/存在检查 |
| 5 | **ARITH** | `d[8]=d[4]-d[8]` 算术运算 |
| 6 | **RETURN** | `return gV`, `throw` 返回/抛出异常 |
| 7 | **ALLOC** | `d[6]=[]`, `d[6][C[8]](...)` push 操作 |
| 8 | **BRANCH** | `d[7]=d[k]?512:425` 条件跳转 |
| 9 | **STRING** | `d[6][C[t]]=d[m]`, `new fh(...)` 正则 |
| 10 | **ALLOC** | 函数参数准备、调用栈创建 |
| 11 | **STRING** | `new fh("\\s",d[5])` 正则匹配 |
| 12 | **STORE** | `P[d[9]]=d[4][C[H]](d[3])` 数据传递 |
| 13 | **CALL** | `C[9][113]=d[9]` 模块初始化 |
| 14 | **STRING** | `d[8]=d[9]+d[m]` 字符串拼接 |
| 15 | **RETURN** | `return EL;` 函数返回 |
| 16 | **ALLOC** | `var r,P,Z,B...` 局部变量声明 |
| 17 | **ALLOC** | `(Z=[])[C[8]](69,T,445)` 静态数组初始化 |
| 18 | **TABLE** | 函数表/类型表初始化 |
| 19 | **EXCEPTION** | `try{for(var RK=x;...` try-catch 循环 |
| 20 | **DOM** | `Is[d[o]]` DOM 操作 |
| 21 | **STORE** | 安全获取全局/对象属性 |
| 22 | **STRING** | `new fh(r,v)` 字符串/正则处理 |
| 23 | **BRANCH** | `try...catch` 安全获取 + 条件跳转 |
| 24 | **CALL** | `W(C[2],null,8,z,FL)` 多参数函数调用 |
| 25 | **EXCEPTION** | `try{...}catch(C){...}` 异常捕获 + 跳转 |

---

## 5. 运行时捕获方案

### 方案 A: Selenium + CDP 原生事件(推荐,成功率最高)

```python
from selenium import webdriver

driver = webdriver.Chrome()

# 注入反检测
driver.execute_cdp_cmd("Page.addScriptToEvaluateOnNewDocument", {
    "source": r"""
        Object.defineProperty(navigator, 'webdriver', {get: () => false});
        Object.defineProperty(navigator, 'plugins', {get: () => [1,2,3,4,5]});
        Object.defineProperty(navigator, 'languages', {get: () => ['zh-CN','zh','en']});
    """
})

# 发送 CDP 原生鼠标事件
driver.execute_cdp_cmd("Input.dispatchMouseEvent", {
    "type": "mousePressed",
    "x": 549.5, "y": 441.2,
    "button": "left", "buttons": 1,
    "clickCount": 1, "pointerType": "mouse"
})
```

### 方案 B: Playwright 无头浏览器

```javascript
const { chromium } = require('playwright');

async function run() {
    const browser = await chromium.launch();
    const page = await browser.newPage();

    // 拦截网络请求
    await page.route('**/api/**', async route => {
        await route.continue_();
    });

    await page.goto('https://target-page.com');

    // 等待 DSL VM 初始化
    await page.waitForFunction(() => {
        return window.AWSCInner &&
               window.AWSCInner._modules &&
               window.AWSCInner._modules['fy'];
    });

    // 执行操作
    await page.mouse.move(500, 400);
    await page.mouse.down();
    // ... 操作序列
    await page.mouse.up();
}
```

### 方案 C: 纯协议验证(成功率极低)

> DSL VM 生成的 token 通常与浏览器上下文强绑定(TLS JA3 指纹、IP、Cookie、请求头等),脱离浏览器后服务端可检测到上下文不匹配。**不建议使用纯协议方案**。

---

## 6. 常见状态码

| Code | 含义 | 处理 |
|------|------|------|
| 0 | **验证通过** ✅ | 取出 sessionId + sig |
| 300 | **风控拦截** | 被拦截,无法通过 |
| 8778 | **验证失败,需重试** | 重试操作 |
| 8776 | **操作太快,需重试** | 增加延迟后重试 |
| 69634 | **通用失败** | 检查参数是否正确 |

---

## 7. Skill 自检清单

- [ ] 我是否完成了 DSL VM 识别(IIFE + 单字母变量 + DG() 解释器)?
- [ ] 我是否提取了变量映射表(`var X=数字`)?
- [ ] 我是否提取了 opcode 列表并分类?
- [ ] 我是否分析了常量表 C[9] 的引用范围?
- [ ] 我是否定位了导出函数注册点?
- [ ] 纯静态分析不够时,我是否尝试了运行时注入方案?

---

Attribution

coco-researchcoco-research
View sourceSee grades on GitHubMore from coco-research →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698461 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →