
Evasion Research
- 31 installs
- 1.6k repo stars
- Updated July 19, 2026
- wgpsec/aboutsecurity
Helps with ai & agent building tasks during AI-assisted development.
About
evasion-research is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.
- evasion-research
- AI & Agent Building
- AI-coding skill
Evasion Research by the numbers
- 31 all-time installs (skills.sh)
- +2 installs in the week ending Jul 27, 2026 (Skillselion tracking)
- Ranked #9,176 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
- Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/wgpsec/aboutsecurity --skill evasion-researchAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 31 |
|---|---|
| repo stars | ★ 1.6k |
| Last updated | July 19, 2026 |
| Repository | wgpsec/aboutsecurity ↗ |
What it does
Helps with ai & agent building tasks during AI-assisted development.
Files
GitHub 安全研究方法论
⛔ 深入参考
- 完整研究流程与命令 → references/research-workflow.md
---
5 步研究流程
1. GitHub 搜索 → gh search repos / gh search code
2. 仓库分析 → gh repo view + 读关键文件
3. 模式提取 → 按 6 大类分类
4. 去重检查 → 查已有知识库避免重复
5. 入库保存 → 写入 evasion-techniques-db.json / loader-components-db.jsonGitHub 搜索命令速查
# 搜仓库
gh search repos "shellcode loader language:C stars:>20" --limit 20
gh search repos "AMSI bypass" --limit 15
gh search repos "syscall direct" --language c --limit 15
# 搜代码
gh search code "VirtualAlloc PAGE_EXECUTE_READWRITE" --language c --limit 30
gh search code "NtCreateThreadEx" --language c --limit 206 大技术分类
| 类别 | 搜索关键词 |
|---|---|
| 内存分配 | VirtualAlloc, HeapCreate, NtAllocateVirtualMemory, MappedFile |
| 代码执行 | CreateThread, EnumWindows, APC, Fiber, callback |
| API 混淆 | API hashing, PEB walk, dynamic resolve |
| 字符串混淆 | XOR, AES, stack strings, compile-time encryption |
| 反分析 | IsDebuggerPresent, anti-VM, sandbox detection |
| Syscall | direct syscall, indirect syscall, SSN, Hell's Gate |
去重规则
| 条件 | 操作 |
|---|---|
| 完全同名 | ❌ 跳过(重复) |
| 同技术不同名 | ❌ 跳过(重复) |
| 同目标不同实现 | ✅ 添加(都有价值) |
| 不同目标类似 API | ✅ 添加(不同用途) |
输出格式
## 研究报告
### 发现的技术
1. [技术名] - [简述] - 复杂度: simple/medium/complex
### 知识库状态
- 新增: X 条
- 重复跳过: Y 条
- 变体补充: Z 条
### 参考链接
- <repository-url>{
"skill_name": "evasion-research",
"evals": [
{
"id": 1,
"name": "research-loader-technique",
"prompt": "需要搜索 GitHub 上新的 shellcode loader 技术并入库。请描述搜索、分析、去重和入库的完整流程。",
"expected_output": "gh search 找仓库 → 分析代码模式 → 去重检查 → 入库新技术",
"expectations": [
"gh search|GitHub搜索|shellcode loader",
"分析代码|模式识别|API调用|技术分类",
"去重|已有技术|避免重复入库",
"入库|记录|技术名称|分类标签"
],
"required_terms": [
"GitHub搜索",
"gh search",
"shellcode loader"
]
},
{
"id": 2,
"name": "research-github-code-search",
"prompt": "你发现现有知识库缺少 direct syscall 的 C 语言实现。你想通过 GitHub code search 找到相关代码。请写出搜索命令,并说明你会如何从搜索结果中筛选出高质量的候选仓库。",
"expected_output": "gh search code \"NtCreateThreadEx\" --language c --limit 20,筛选标准:星数>100、近期更新、无恶意特征、代码可复现",
"expectations": [
"gh search code|GitHub代码搜索|code search",
"NtCreateThreadEx|direct syscall|搜索关键词",
"stars|星数|质量筛选|活跃度",
"--language c|--limit|搜索参数",
"恶意检测|可疑仓库|排除|筛选标准"
],
"required_terms": [
"gh search code",
"NtCreateThreadEx",
"--language c"
]
},
{
"id": 3,
"name": "research-deduplication-decision",
"prompt": "你通过 GitHub 研究发现了两项 AMSI 绕过技术:A) AmsiScanBuffer 内存字节 patch(已有类似技术入库),B) 使用 AES 加密绕过 AMSI 特征检测(新目标新实现)。请判断哪些应该入库,哪些应该跳过,理由是什么?",
"expected_output": "A 跳过(已有同类技术入库),B 入库(新目标不同实现有价值)。入库判断:同名技术跳过,同技术不同名跳过,同目标不同实现或不同目标类似 API 都应入库",
"expectations": [
"同名技术|跳过|重复|不入库",
"同技术不同名|跳过|重复",
"同目标不同实现|入库|有价值|补充",
"不同目标类似API|入库|不同用途",
"判断标准|入库规则|决策"
],
"required_terms": [
"AmsiScanBuffer",
"跳过",
"入库"
]
}
]
}
{
"skill_id": "evasion-research",
"recall_tests": [
{
"id": 1,
"type": "keyword_positive",
"description": "核心关键词",
"keywords": [
"evasion research",
"loader",
"technique"
]
},
{
"id": 2,
"type": "keyword_positive",
"description": "知识库搜索",
"keywords": [
"github",
"knowledge-base",
"shellcode"
]
},
{
"id": 3,
"type": "keyword_negative",
"description": "不应被recon召回",
"keywords": [
"subdomain",
"port scan"
]
}
],
"llm_tests": [
{
"id": 1,
"name": "evasion-research-scenario",
"scenario": "需要研究目标 EDR 的检测机制以制定绕过策略。请搜索规避检测研究方法论。",
"max_rounds": 2,
"expect_tool_calls": [
{
"tool": "list_skills",
"keyword_contains": "evasion|research|绕过|检测"
},
{
"tool": "read_skill",
"id": "evasion-research"
}
]
}
]
}
研究流程详细参考
提取自 Evasion-SubAgents/research.md,供 references/research-workflow.md 引用。
Step 1: GitHub 搜索
使用 gh CLI 搜索。先宽后窄。
# 搜索仓库
gh search repos "shellcode loader language:C stars:>20" --limit 20
gh search repos "AMSI bypass" --limit 15
gh search repos "syscall direct" --language c --limit 15
# 搜索代码模式
gh search code "VirtualAlloc PAGE_EXECUTE_READWRITE" --language c --limit 30Step 2: 仓库分析
对有价值的仓库,获取关键文件:
# 查看仓库信息
gh repo view owner/repo
# 列出文件
gh api repos/owner/repo/contents
# 获取文件内容
gh api repos/owner/repo/contents/path/to/file.c --jq '.content' | base64 -dStep 3: 模式提取
按以下类别分类提取的技术:
| 类别 | 关键词 |
|---|---|
| 内存分配 | VirtualAlloc, HeapCreate, NtAllocateVirtualMemory, MappedFile |
| 代码执行 | CreateThread, EnumWindows, APC, Fiber, callback |
| API 混淆 | API hashing, PEB walk, GetProcAddress, dynamic resolve |
| 字符串混淆 | XOR, AES, stack strings, compile-time encryption |
| 反分析 | IsDebuggerPresent, CheckRemoteDebugger, anti-VM, sandbox |
| Syscall | direct syscall, indirect syscall, SSN, Hell's Gate |
Step 4: 去重检查
入库前必须去重:
| 条件 | 操作 |
|---|---|
| 完全同名 | SKIP — 重复 |
| 同技术不同名 | SKIP — 重复 |
| 同目标不同实现 | ADD — 两者都有价值 |
| 不同目标类似 API | ADD — 不同用途 |
| 同源码同方法 | SKIP — 重复 |
| 同源码不同方法 | ADD — 变体 |
Step 5: 输出总结
研究完成后提供:
1. 发现的技术: 列表 + 简述 2. 复杂度评估: simple / medium / complex 3. 知识库状态: NEW / DUPLICATE / VARIATION 4. 参考链接: GitHub URL
Bash 命令规范
✅ 正确:
gh search repos "shellcode loader"❌ 错误:
gh search repos "query" 2>/dev/null # 不要重定向
cd "/some/path" && gh search repos ... # 不要 cd 组合Related skills
AI & Agent Buildingagents