本文へ移動
cccskills
無料GitHub で公開

web-attack

HTTP/Web exploitation until a server shell or equivalent OS execution. Use after /web-recon has mapped the application. Covers injection, upload, LFI, SSRF/XXE, SSTI, deserialization, JWT/SAML, API logic, desync/cache/parser, and Web-controlled backend abuse. WAF stays here. Endpoint blocks after OS execution hand off to /edr-bypass then return. Direct non-HTTP service ports belong to /recon or /service-attack.

インストール方法を見る

含まれるファイル(25)

  • SKILL.md5.3 KB
  • references/api-security.md20.3 KB
  • references/auth-bypass.md1.9 KB
  • references/cmdi.md5.0 KB
  • references/deserialization.md1.6 KB
  • references/http-desync.md5.2 KB
  • references/java-deserialization.md3.3 KB
  • references/jwt.md1.4 KB
  • references/lfi-rfi.md4.0 KB
  • references/orm-leak.md5.3 KB
  • references/parser-differentials.md4.4 KB
  • references/saml-security.md3.6 KB
  • references/server-side-client-abuse.md3.3 KB
  • references/sqli-mssql-chain.md6.3 KB
  • references/sqli-mysql-chain.md5.9 KB
  • references/sqli-oracle-chain.md4.7 KB
  • references/sqli-postgresql-chain.md4.6 KB
  • references/sqli.md8.0 KB
  • references/ssrf.md4.6 KB
  • references/ssti.md5.3 KB
  • references/upload.md3.6 KB
  • references/waf-bypass.md6.8 KB
  • references/web-cache.md4.1 KB
  • references/xss.md3.8 KB
  • references/xxe.md1.5 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

/web-attack — Web 漏洞利用

scope: 从 HTTP/Web 输入把当前链打到 服务器 Shell(或等价 OS 执行)。直连 1433/3306/6379 不是这里。WAF/上传过滤留本模块。▸ Shell 一旦拿到,记 notes;候选只认 ~/.claude/skills/shared/modules.yaml(modules.py tail)。

工具:../shared/tools.md。产品 CVE 由 /web-recon 出;利用仍在本模块打到 shell。有精确版本才读 ../shared/cve-enrichment.md。


开局与收尾

开局第一件事:Read ./notes.md。没有则 python ~/.claude/skills/bin/notes.py init。只按已拿下/凭据继续。 走到哪条链,才 Read 一份 references/<file>.md。禁止开局全读、禁止凭记忆写 payload。 监听 / sudo / 输密码 / Permission denied:立刻停,说明等操作者做什么,等回报。不要假装已成功。 收尾:

  1. 追加 ./notes.md
  2. python ~/.claude/skills/bin/modules.py tail <本模块名> Read 备用:~/.claude/skills/shared/modules.yaml 禁止 ./modules.yaml 和 python ../bin/...
  3. 优先 default_next;never_default 不得当作默认(操作者点名除外)
  4. 名册外的名字不许建议
  5. 停。等操作者选 /模块 或 /clear /edr-bypass 半条链未完:打通后回本模块,不要 /clear。

0. 成功条件

Web 输入 / 功能 / API
        ↓
漏洞成立
        ↓
命令执行 / 可执行文件落地 / 管理身份 + 执行点
        ↓
服务器 Shell?
├─ NO  → 继续当前 Web 链(SQLi→OS、XSS→后台→上传、SSRF→Redis)
└─ YES → /web-attack COMPLETE

不算: sqlmap 报 injection、alert(1)、能传 jpg、Nuclei MATCH。


0.4 不要吸收

东西去哪
直连数据库/Redis 端口/recon /service-attack
路径枚举 / 指纹 / 已知 CVE 发现/web-recon
本机提权/privesc-*
JWT 弱密钥短字典本模块 jwt_tool -C,爆了立刻签发继续打 shell
JWT 长时间 GPU / 全 rockyou仅操作者另开凭据作业 → 候选 /creds

0.5 EDR

WAF / 上传过滤 / parser / cache = 本模块。
OS 上执行被 AV 杀 → /edr-bypass → 回本模块拿完 shell。


1. 决策树(按手上是什么,只读一份)

HTTP 参数像注入
  → SQLi:先 Read references/sqli.md 指纹 DBMS
     再只读 sqli-mssql|mysql|postgresql|oracle-chain.md 之一
  → CMDi:Read references/cmdi.md
  → SSTI:Read references/ssti.md
  → NoSQL/ORM:Read references/auth-bypass.md 或 references/orm-leak.md
文件 / 包含 / 上传
  → 上传:Read references/upload.md(先问栈,不要默认 PHP)
  → 包含:Read references/lfi-rfi.md
服务端出站
  → XXE:Read references/xxe.md ;SSRF:Read references/ssrf.md ;Rogue MySQL:Read references/server-side-client-abuse.md
Token / SAML / JWT
  → JWT:Read references/jwt.md ;SAML:Read references/saml-security.md
序列化
  → PHP/.NET/pickle:Read references/deserialization.md
  → Java/Fastjson:Read references/java-deserialization.md
代理/CDN(recon 已确认才走)
  → desync / cache / parser / waf-bypass 各读当前那一份
API 逻辑
  → Read references/api-security.md ;XSS 接执行 Read references/xss.md

同一条链打穿再换。不要按课表把上表全跑一遍。


2. 注入类

sqlmap -r request.txt -p param --current-user --current-db --privileges --batch

确认 DBMS 后再 Read 对应 references/sqli-*-chain.md。不要一上来 --os-shell。


3. 文件 / 包含 / 上传

上传 → 现在 Read references/upload.md:定位 URL + 解释器执行 = getshell。IIS / Tomcat / Node,禁止默认 .phtml。

LFI → 现在 Read references/lfi-rfi.md。能 include 才走 filter chain。


4. 出站

XXE / SSRF 分别 Read。SSRF 拿到云密钥、Web 再走不动 → 候选 /cloud-recon。还能 gopher/Redis 拿 shell 就先打完 Web。


5. JWT / SAML

弱密钥 本模块砸:

jwt_tool <token> -C -d /usr/share/wordlists/rockyou.txt
hashcat -m 16500 jwt.txt wordlist.txt
jwt_tool <token> -S hs256 -p '<密钥>' -T

完整算法绕过现在 Read references/jwt.md。SAML 现在 Read references/saml-security.md。


6. 反序列化

头:PHP O:、Java rO0AB、.NET AAEAAAD、pickle \x80。按栈只读一份。


7. API / XSS

越权不是终点,接到执行。现在 Read references/api-security.md 或 references/xss.md。


8. 代理边界

仅 recon 已确认 CDN/反代时,才 Read desync/cache/parser/waf 中当前那一种。


9. 完成后

shell: host / user / 怎么拿到的
RESTORE: 删的马 / 还原的 web.config

写入 ./notes.md(shell 用户/路径/RESTORE)。候选只按「开局与收尾」跑 python ~/.claude/skills/bin/modules.py tail。

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

ad-attack

無料

Active Directory exploitation after domain credentials exist: Kerberos (AS-REP/Kerberoast including cracking the ticket then using the account), delegation, NTLM coercion/relay, lateral movement, ACL abuse, ADCS ESC1-ESC17 and CVE paths, dMSA/BadSuccessor, Kerberos reflection, identity confusion, management-plane, domain trust, and domain persistence. Own the current AD chain through crack-and-use to DA, equivalent domain control, or the targeted host SYSTEM. Do not stop after requesting TGS. Host C2 belongs to /post. Operator chooses next modules.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

ad-recon

無料

Active Directory reconnaissance with or without credentials: user/group/computer enumeration, ACL/delegation, ADCS, modern Windows LAPS, BloodHound, Server 2025/dMSA/Ghost SPN candidates, and trust mapping. Recon only — do not exploit Kerberoast-to-DA, DCSync, or change passwords. Operator may select /ad-attack after cards are ready.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

cicd

無料

CI/CD pipeline and software-supply-chain exploitation: Jenkins, GitHub Actions, GitLab CI/CD, Azure DevOps, Gitea/Forgejo, self-hosted runners/agents, poisoned pipeline execution, artifact/cache abuse, dependency confusion, third-party Action trust, GitOps/registry poisoning, workload identity/OIDC, and emerging agentic CI/CD. Use this skill when the operator has access to a CI/CD system, source repository, build/deploy configuration, runner/agent, artifact/package/registry path, or software delivery trust chain.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

Cloud control-plane exploitation for AWS, Azure/Entra, GCP, and Alibaba Cloud: IAM/RAM privilege escalation, impersonation, cross-account trust, serverless/compute control, and cloud-native persistence. Host OS persistence/C2 after root/SYSTEM belongs to /post. K8s RBAC belongs to /k8s. Operator chooses next modules; new identities default to /cloud-recon first.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

Cloud control-plane reconnaissance for AWS, Azure/Entra, GCP, and Alibaba Cloud: identity, IAM/RAM, trust, resources, metadata, and managed-Kubernetes cloud-side boundary. Recon only — no policy changes, no privilege escalation. Operator may select /cloud-attack or /k8s.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

creds

無料

Credential operations: secret discovery, classification, extraction, conversion, offline cracking of hashes the operator already has as a credential job, policy-aware spraying, NetNTLM capture, generic SMB relay, and Windows/Linux harvest. Do not hijack an in-progress /ad-attack Kerberoast/AS-REP chain (that module cracks and uses the ticket itself). Do not DCSync, read LAPS LDAP, or escalate cloud IAM. Usable credentials are recorded; the operator chooses the next module.

日本語の概要は準備中です。原文の説明を表示しています。

pale-knight/redteam-skill492026年8月23日 更新

pale-knight のスキルをすべて見る

このスキルの問題を報告する