Back to skills
SKILL.md
Threat Hunting
ASecurityUse for blue-team threat hunting, detection engineering with Sigma/YARA, SIEM query design, and incident detection validation.
- 2 stars
- 0 votes
- 0 copies
- 0 views
- Added October 17, 2026
Works with
Security analysis
100/100Pro scans all 3 files and shows the line behind each finding
npx -y skills add NetVar1337/agent-skill-canon --skill threat-hunting --agent claude-codeAre you the author of Threat Hunting?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/netvar1337-threat-hunting-3eafba03)---
name: core-threat-hunting
description: "Use for blue-team threat hunting, detection engineering with Sigma/YARA, SIEM query design, and incident detection validation."
version: 1.0.0
license: MIT
metadata:
package: unleash-skills
author: NetVar1337/unleash
category: reverse-skill
upstream: https://github.com/zhaoxuya520/reverse-skill
---
> Bundled with Unleash skills pack. Source: https://github.com/zhaoxuya520/reverse-skill
# Threat Hunting & Detection Engineering
## ACTION REQUIRED(读完后立刻执行)
1. `NOW`: 确认蓝队/狩猎授权与数据源范围(SIEM、EDR 导出)
2. `NOW`: 明确假说(hypothesis)再查数,避免无脑刷告警
3. `NEXT`: 工具与数据接入方式
4. `ACT`: 假说 → 查询 → 验证 → 规则化
## 适用场景
- 威胁狩猎(hypothesis-driven)
- Sigma / YARA 检测工程
- 告警调优、误报分析
- 与 `malware-analysis/`:样本侧 IOC → 本 skill 落地检测
- 与 `digital-forensics/`:案件伪影 → 横向狩猎
## 工作流
### 1. 建假说
```text
例:攻击者用 living-off-the-land 做横向
→ 数据源:Sysmon 1/3/10、Windows Security 4624/4648
→ 成功标准:发现异常父进程或罕见账户日志源
```
### 2. 查询与堆叠
```text
□ 基线:正常管理员行为时段与主机
□ 异常:新服务、编码 PowerShell、异常出站
□ 关联:同账号多主机短时登录
```
### 3. 规则化
```yaml
# Sigma 骨架见 malware-analysis;本 skill 强调:
# - 误报面
# - 数据源字段映射
# - 响应 playbook 链接
```
### 4. 验证
```text
□ 原子测试(Atomic Red Team)仅在授权实验室
□ 回放历史日志验证召回
```
## 工具链
| 工具 | 用途 |
|------|------|
| Sigma CLI / sigmac | 规则转换 |
| YARA | 文件/内存 |
| SIEM(ELK/Splunk 等) | 查询 |
| osquery | 端点狩猎 |
| Atomic Red Team | 检测验证(实验室) |
## 参考
- `references/hunting-loop.md`
- `../malware-analysis/references/yara-sigma-rules.md`
- `../digital-forensics/`
## 路由上下文
**上游**: MASTER R27
**下游**: 确认入侵 → forensics;恶意样本 → malware-analysis
**MUST NOT**: 在无授权生产环境跑攻击模拟
## 任务完成自检
- [ ] 是否有明确假说与结论?
- [ ] 规则是否注明误报与数据源?
- [ ] Checklist?
Files in this skill
- SKILL.md
- agents/openai.yaml
- references/hunting-loop.md
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.