DeepSeek Harness plugin

dsh-htb-skills

HTB (Hack The Box) pentesting skill library plugin for DeepSeek Harness: 112 distilled technique cards (T1 routing / T2 deep-dive / T3 reference tiers, disable-model-invocation for hidden tiers) plus

Jump to install

Source facts

Repository
qingsiweisan/dsh-htb-skills
Latest update
Aug 20, 2026
Category
Models & Providers
GitHub stars
0
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/qingsiweisan/dsh-htb-skills
Plugin: dsh-htb-skills
Author: qingsiweisan

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer3 files
README.mdSource · read only

dsh-htb-skills

HTB(Hack The Box)打靶技能库,封装为 DeepSeek Harness 插件。

这是一个社区项目,并非 DeepSeek 官方插件,也不代表 DeepSeek 的认可或背书。

是什么

  • 114 张技能卡:从 Reasonix Studio 的 HTB 技能库移植,覆盖 Web / AD-Windows / Linux / 数据库 / 云 / 取证等 10 个领域;
  • 插件形态:包内自带 skills/ 树,通过官方 @deepseek-ai/dsh-skill-filesystem 的 provider 机制注册为名为 htb 的技能提供者——技能随包版本化、git 管理,编辑即时热重载,不写任何文件进 ~/.dsh/skills
  • 三层结构:T1 路由卡进会话目录;T2 深度卡 / T3 参考卡带 disable-model-invocation: true 隐藏,按名精确加载;htb-skill-index 是全部卡名的领域×层级总索引。

安装

# 1. 克隆本仓库
git clone https://github.com/qingsiweisan/dsh-htb-skills.git

# 2. 安装到 DSH profile(link 安装,本地改代码即时生效)
dsh plugin --profile web add link:./dsh-htb-skills
#   或从 npm 安装:
#   dsh plugin --profile web add dsh-htb-skills

# 3. 重启 dsh web,新建会话即可看到 htb-skill-index 等 T1 卡

Linux / Kali 同理,路径换成克隆目录即可。

> 注意:插件挂载后,~/.dsh/skills 里的同名散装技能会与插件里的卡重复(同名时 preset 层会遮蔽全局层)。迁移时先备份再清空散装目录: > > ``sh > mv ~/.dsh/skills ~/.dsh/skills.bak # 确认插件生效后再删除 > ``

目录结构

dsh-htb-skills/
├── package.json          # dsh.bundle.patch 指向 cordis.patch.yml
├── cordis.patch.yml      # 注入 host 插件行:- id: htb-skills
├── src/index.js          # 插件本体:注册 htb skill provider(无浏览器半区)
├── skills/               # 114 张技能卡(SKILL.md + frontmatter)
│   ├── htb-skill-index/  #   总索引(T1,进目录)
│   ├── box-startup/      #   T1 路由卡示例(含 key-state 产物规范)
│   └── ...               #   T2/T3 卡带 disable-model-invocation: true
├── docs/
│   ├── CARD-NORMS.md     # 卡片写作规范(可解析工件/漏洞形状/语义 rubric/子代理隔离)
│   └── ARCHITECTURE.md   # 架构与变异点地图(机制域 vs 内容域,改框架先读它)
└── scripts/
    ├── triage_skills.py  # 分层/修 whenToUse/修坏链/重生成索引卡(末尾自动跑路由审计)
    ├── fix_yaml.py       # 全量 YAML 引号规范 + PyYAML 校验(幂等:解码后重转义)
    ├── audit_routing.py  # 路由完整性审计(只读):名字唯一性/MAPPING 双向一致/
    │                     #   tier 与 disable-model-invocation 一致/索引卡新鲜度/跨卡引用可解析
    └── verify_run.py     # 判定层:flag/cred 捕获事件的确定性复核(读磁盘 transcript)

卡片的 frontmatter 约定

---
name: web-attacks
description: 'Web 攻击综合手册:OWASP 2025 映射→…'
whenToUse: '目标有 HTTP/HTTPS 攻击面时:…'
# T2/T3 卡额外带这一行;T1 卡不写:
disable-model-invocation: true
metadata: { domain: web, tier: T1 }
---
  • domain 取值:meta / web / ad-win / linux / db / cloud / creds / forensics / network / tools
  • tier 取值:T1(进目录)/ T2(深度,按名加载)/ T3(参考与题源,按名加载)

维护

改完技能后跑一遍校验链(需 pip install pyyaml;三个脚本都幂等,可重复跑):

python scripts/fix_yaml.py        # 规范引号 + 校验全部 frontmatter
python scripts/triage_skills.py   # 分层元数据 + 重生成索引卡(末尾自动跑路由审计)
python scripts/audit_routing.py   # 路由审计:引用解析/索引一致性(任一错误退出码 1)

audit_routing.py 检查的路由面:卡名合法且唯一、目录名与卡名一致(警告)、MAPPING 与磁盘双向覆盖(索引不会挂空、不会漏卡)、metadata.tierdisable-model-invocation 一致、索引卡与 MAPPING 生成内容逐字节一致(STALE-INDEX)、卡内 backtick/[[wiki]]/中文路由关键词(加载/详见/见/查…)引用的名字全部可解析(DANGLING-REF),以及引号翻倍回归守卫(QUOTE-RUN-CORRUPTION)。

新增一张卡 = 在 skills/ 下建目录 + 写 SKILL.md + 在 scripts/triage_skills.pyMAPPING 里登记 domain/tier,重跑校验链。插件 watch 默认开启,编辑后无需重启。

发布到 GitHub / npm

git init
git add .
git commit -m "dsh-htb-skills: initial HTB skill library plugin"
gh repo create dsh-htb-skills --public --source=. --push   # 或手动 git remote add + push
  • 建议给仓库加 dsh-plugin topic 方便被发现(官方生态指引);
  • 发布 npm:npm publish,之后两台机器都可以 dsh plugin --profile <name> add dsh-htb-skills 安装;
  • 升级 = git pull(link 安装)或 dsh plugin --profile <name> update dsh-htb-skills(npm 安装)。

兼容范围

  • DeepSeek Harness 0.1.0-rc.x(与 @deepseek-ai/dsh-skill-filesystem ^0.1.0-rc.6 对齐);
  • Node ^22.19.0 || >=24.0.0
  • Harness 仍处于开发者预览期,升级后如有破坏性变更请对照上游 skill-filesystem 的导出再使用。

许可证

MIT。技能卡内容移植自 Reasonix Studio 的个人技能库。