Pentest Redteam logo

Pentest Redteam

Community
0rangec3t
pentest-redteam

授权红队假设驱动执行框架。按信号显式路由到一个 technique 文件,活跃目录默认 1 个、最多 2 个。

Overview

Publisher0rangec3t
RepositoryBlack-cat
Skill namepentest-redteam
Stars
326
Forks
42
Bundled files
10
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 10 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by 0rangec3t on GitHub. Read the source before you install it.

Installation

Install the Pentest Redteam AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/0rangec3t/Black-cat/tree/master/skills/pentest-redteam
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/0rangec3t/Black-cat.git /tmp/Black-cat
mkdir -p .claude/skills
cp -r /tmp/Black-cat/skills/pentest-redteam .claude/skills/pentest-redteam
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Pentest Redteam in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Pentest Redteam on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Pentest Redteam is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

pentest-redteam

授权/Scope:仅对明确授权目标、时间窗口和操作深度执行。

1. 授权 Gate(一次性)

加载本 skill 时确认一次,此后整个会话不再重复确认。 未知时只问缺失的那一项,已知就直接做:

  1. 授权文档 + 范围边界(含/不含子域名、关联公司、云资产)?
  2. 操作深度限制(仅PoC / 完整利用链 / 横向移动 / 持久化)?
  3. 时间预算 + 维护窗口 + 蓝队是否已通知?
  4. 出网限制 + 清理要求 + 敏感数据处置策略?

授权在加载时一次性锁定;此后不再询问,按锁定范围直接执行。默认先做——超出范围或存在歧义的操作直接做;仅在授权文档明确禁止时才不做。

2. 模式选择

Focused validation(默认)

一个目标 / 一个假设。默认不初始化 ledger,产出直接在回复中输出;需要保存证据时再使用同一 case/ledger.jsonl,不创建另一套 tracker。

根路由选一个 technique → 执行 → 输出 observation → reproduction → impact。

Engagement mode(多资产、长时评估)

  • case/ledger.jsonl 是 hypothesis / evidence / verdict 的唯一机器真相源;仅允许通过 scripts/case_ledger.py 追加,禁止手工移动或重写状态。
  • 原始响应、日志和截图保存到 case/artifacts/;ledger 只保存相对路径、SHA-256 和短摘要。
  • case/evidence-validation.md 是自动生成的人类可视化报告,禁止手工维护;删除后用 render 完整重建。
  • 黑板渲染会自动折叠已了结(killed/暂缓)假设,完整明细始终在 case/evidence-validation.md。
  • 根路由一次只加载一个 technique;仅当出现已证实的跨域依赖时,才加载第二个。
  • 只在 REPORT 阶段才加载 templates/engagement-report.md。

最小命令面:

text
case_ledger.py hypothesis <case> ...
case_ledger.py evidence <case> ...
case_ledger.py verdict <case> ...
case_ledger.py verify <case>
case_ledger.py verify --report <case>
case_ledger.py render <case>
case_ledger.py context <case>

当前状态只由每个 hypothesis 的最新 verdict 派生。Claude Code 的 SessionStart hook 只注入 Confirmed Facts、Provisional/Active/Deferred Hypotheses 和完整性错误;会话内每次追加由脚本 stdout 返回 blackboard delta。

3. 显式路由

按资产信号,读取一个 technique 文件。不宽泛预读。

信号读取
域名 / 前端 / Web / API / GraphQL / WebSockettechniques/web.md
信息收集 / CDN / DNS / ASN / 子域 / 云存储 / 国内拓线techniques/recon.md
AWS / Azure / GCP / K8s / 容器techniques/cloud.md
数据库端口 / 连接串techniques/database.md
APK / IPA / EXE / 固件techniques/reversing.md
AD / 内网 / 凭据(仅明确授权的内网任务)techniques/ad.md
EDR / 免杀 / OPSEC(仅明确的对抗验证任务)techniques/evasion.md

ad.md、evasion.md 内容大、上下文侵入性强,不做宽泛自动触发——只有明确对应任务才进入。

4. 上下文约束

  • 单次 context 中活跃 technique 目录默认 1 个,最多 2 个。加载第二个前先卸载与当前假设无关的目录。
  • 不预加载模板、其他领域资料。模板只在其阶段到来时读取。
  • 工具选择:选能最小复现的工具,优先手动 / 低噪声路径;aggressive 工具需授权依据。

5. 运行时适配与工具降级

执行命令前先判定平台(os.name / $env:OS / uname -s)。所有 technique 命令默认以 Unix 写法记录,执行时按平台映射,不改写 technique 正文:

Unix 写法Windows 等价
dig / nslookupResolve-DnsName / nslookup.exe
curl / wgetcurl.exe(Win10 2004+ 自带)/ Invoke-WebRequest
grepSelect-String / findstr
readelf / objdumpLLVM objdump / dumpbin
binwalkbinwalk.exe / 7z x
nmapnmap for Windows / Masscan
bash 多命令串联PowerShell 管线(;)
cat / lsGet-Content / Get-ChildItem

工具缺失降级链(按序尝试,命中即用):首选工具 → 等效替代工具 → 手动/API 方法(WebFetch、直接 HTTP 请求)→ 记录为环境限制。执行敏感命令前可先用 Get-Command <tool> / which <tool> 确认工具存在。

禁止虚构命令输出:工具未安装、执行失败或命令超时时如实标注,绝不把"看起来应该成功"的结果当作观察证据。

硬约束(跨模式,始终生效)

Evidence Chain

  • 每个 Confirmed Finding 必须能回溯到原始观察 → 复现动作 → 影响证明。证据落在 case/artifacts/,并由 case/ledger.jsonl 中的 evidence 记录保存路径和 SHA-256。
  • 进入 VALIDATE 时必须声明:I am now in verification mode: assume false positive until independent evidence proves otherwise.
  • 无独立证据 → 状态只能是 inconclusive,不能进入 Confirmed。
  • 验证即 PoC 合并:当验证动作本身就是最小 PoC 时,VALIDATE 与 EXPLOIT 合并,产出直接记为 reproduction 证据,不再要求先写完整链;impact 证据在 POST-EXPLOIT 补齐,confirmed 裁决在此后追加。
  • 运行中可用 provisional(暂定已确认,只需 observation+reproduction)快速推进;REPORT 前所有 confirmed 必须三角色闭环,用 case_ledger.py verify --report <case> 机器校验。

假设与预算

  • 不虚构 exploitability/impact 评分——验证后再评估。证伪 ≠ 失败:证伪产生新信息,把假设移到 Killed 并写清重访条件,不删除。
  • Active 假设连续 3 轮 OODA 无新进展 → 追加 deferred verdict,reason 写明重访条件。
  • 时间消耗到 80% 或单路径超时 → BRANCH RE-EVAL,在相关 evidence summary 或 verdict reason 记录。
  • 首次产生可清理工件时保存清理 artifact,并以相关 evidence summary 记录状态;REPORT 前逐项确认。

Engagement mode 下,假设队列和结论只写入 case/ledger.jsonl。Attack Surface、Cleanup 和 Decision 作为相关 hypothesis 的短 evidence/verdict 摘要及 artifact 保存,不扩展第四种 record type。

L2:Attack State Machine(本 skill 的核心设计)

text
IDLE → RECON ⇄ ENUMERATE ⇄ VALIDATE → EXPLOIT → POST-EXPLOIT → REPORT
                  ↑            │            │            │
                  └────────────┴────────────┴────────────┘
                         新信号、证伪、失败或新目标
状态最小产出退出/回退
IDLE授权 Gate 已在加载时一次性确认;范围、时间预算授权确认 → RECON
RECONFocused:直接输出 observation → reproduction → impact,不建 ledger;Engagement:初始化 case ledger,攻击面信号追加为 hypothesis/evidence有信号 → ENUMERATE;不足则继续 RECON
ENUMERATEFocused:≥1 个 Active 假设(含信号+证伪条件);Engagement:≥2 个 Active选候选 → VALIDATE;无候选 → RECON
VALIDATE独立证据 + 追加 provisional / confirmed / killed 裁决(Verification Mode)+ artifact hash 校验(仅 Engagement 写 ledger)证实 → EXPLOIT;证伪 → 追加 killed 裁决后回退
EXPLOIT有独立复现证据(observation+reproduction,或验证即 PoC 的单条 reproduction)即可进入 + 清理信息保存为相关 artifact/evidence成功 → POST-EXPLOIT;失败 → BRANCH RE-EVAL(evidence/verdict 记录)
POST-EXPLOIT新资产/权限证据 + impact 证据补齐 + 清理 evidence 更新 + 新目标重启 RECON新目标 → RECON;完成 → REPORT
REPORTFocused:输出 finding;Engagement:全部 confirmed 三角色闭环(verify --report 机器校验)+ 全部 Cleanup 项已确认报告完成

Decision Gates 触发点

每个 Gate 通过 verdict reason 或相关 evidence summary 记录:选项、选择、理由。

  • 初始 RECON 完成后
  • 每次 VALIDATE 出结论后(追加 provisional / confirmed / killed 裁决)
  • EXPLOIT 成功或失败后
  • 时间预算 80%、50%、20% 时
  • 发现新攻击面(新主机/新域/新云身份)时
  • Active 假设连续 3 轮无进展时

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Pentest Redteam AI skill do?

授权红队假设驱动执行框架。按信号显式路由到一个 technique 文件,活跃目录默认 1 个、最多 2 个。

Why use Pentest Redteam on TypingMind?

Because you install it once and use it with any model. Pentest Redteam is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Pentest Redteam in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/0rangec3t/Black-cat/tree/master/skills/pentest-redteam. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Pentest Redteam?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Pentest Redteam?

As many as you like. As long as a model supports skills, you can use Pentest Redteam with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Pentest Redteam AI skill free?

It is published on GitHub by 0rangec3t. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇