中文文档 | English

AboutSecurity

Penetration testing knowledge base with security methodologies in AI Agent-executable format.

Core Modules

Skills/ — 180+ skill methodologies covering the full chain from recon to post-exploitation

• ai-security/ — AI security (prompt injection, model jailbreaking, prompt leaking, agent attack chains)
• cloud/ — Cloud environments (Docker escape, K8s attack chains, AWS IAM, Alibaba Cloud, Tencent Cloud, Serverless)
• code-audit/ — Code auditing (PHP 8-skill system, Java 8-skill system covering injection/file/serialization/auth/framework/exploit chains)
• ctf/ — CTF competitions (Web challenges, reversing, PWN, cryptography, forensics, AI/ML)
• dfir/ — Digital forensics & incident response (memory forensics & anti-forensics, disk forensics, log evasion)
• evasion/ — Evasion techniques (C2 frameworks, shellcode generation, security research)
• exploit/ — Exploitation (organized by subcategory)
  • advanced/ — Advanced exploitation (HTTP smuggling, race conditions, supply chain attacks, OT/ICS, crypto attacks)
  • auth/ — Authentication & authorization (JWT, OAuth/SSO, IDOR, CORS, CSRF, cookie analysis)
  • binary/ — Binary exploitation methodology and tools
  • network-service/ — Network service pentesting by port/protocol (SMB, FTP, SMTP, DNS, LDAP, etc.)
  • web-method/ — Web methodology (injection, XSS, SSRF, SSTI, file upload, deserialization, WAF bypass...)
• general/ — General (report generation, supply chain auditing, mobile backend API)
• hardware/ — Hardware/physical access pentesting
• lateral/ — Lateral movement (AD domain attacks, NTLM relay, database pivoting, Kerberoasting, ACL abuse)
• malware/ — Malware (sample analysis methodology, C2 beacon config extraction, sandbox evasion)
• mobile/ — Mobile app pentesting (Android, iOS)
• postexploit/ — Post-exploitation
  • post-exploit-linux/ / post-exploit-windows/ — OS-level privilege escalation, credential theft
  • persist-maintain/ — Persistence techniques (cron, services, webshell)
  • tool-delivery/ — Tool delivery to compromised hosts
  • product/ — Product-specific post-exploitation tactics (ArgoCD, Harbor, databases, middleware, Portainer, RabbitMQ)
• recon/ — Reconnaissance (subdomain enumeration, passive information gathering, JS API extraction)
• threat-intel/ — Threat intelligence (IOC evasion, APT simulation, threat hunting evasion)
• tool/ — Tool usage (fscan, nuclei, sqlmap, msfconsole, ffuf, hashcat)

Dic/ — Dictionary library (lowercase hyphen-separated naming, each directory has _meta.yaml metadata)

• auth/ — Usernames/passwords (complexity-rule passwords, WPA, pinyin names)
• network/ — DNS servers, excluded IP ranges
• port/ — Service-specific brute-force dictionaries (mysql, redis, ssh, etc. — 19 types)
• regular/ — General-purpose dictionaries (numbers, letters, addresses, keywords)
• web/ — Web directories, API parameters, middleware, upload bypass, webshells, HTTP headers

Payload/ — Attack payloads (lowercase hyphen-separated naming, each directory has _meta.yaml metadata)

• sqli/, xss/, ssrf/, xxe/, lfi/, rce/, upload/, cors/, hpp/, format/, ssi/, email/, access-bypass/, prompt-injection/

Vuln/ — 600+ vulnerability entries, structured vulnerability data organized by product

• ai/ — AI-related (ComfyUI, Dify, LangFlow, AnythingLLM, etc.)
• cloud/ — Cloud platforms (AWS API Gateway, etc.)
• middleware/ — Middleware (ActiveMQ, Nacos, Grafana, Jenkins, RocketMQ, etc. — 394 entries)
• network/ — Network devices (routers, switches, etc.)
• web/ — Web applications (1Panel, WordPress, OFBiz, etc.)

Tools/ — Declarative tool configurations for programmatic orchestration frameworks

• scan/, fuzz/, osint/, poc/, brute/, postexploit/
• See Tools/README.md for details

Tools/ vs skills/tool/: YAML files under Tools/ are structured interface definitions for programmatic tool orchestration frameworks (parameter types, command templates, output parsers), suitable for automation engines. SKILL.md files under skills/tool/ are natural language methodologies for LLM Agents (when to use, how to choose parameters, how to interpret results). If you only use LLM Agents like Claude Code, focus on skills/tool/.

postexploit/ vs Vuln/: Skills under postexploit/ are the post-exploitation layer — what to do after gaining access (privilege escalation, persistence, credential extraction, lateral movement, product-specific tactics). Entries under Vuln/ are the vulnerability data layer — affected versions, PoC code, specific exploitation steps per CVE. In short: Skills tell you "what to do after you're in", Vuln tells you "how to get in".

Quick Start

1. Clone the Repository

git clone https://github.com/wgpsec/AboutSecurity.git

2. Sync Skills to Your Project

# Sync all security skills to your working project
cd AboutSecurity
./scripts/sync-claude-skills.sh --target /path/to/your-project

# Result: creates .claude/skills/<skill-name>/ symlinks in the target project
# Claude Code will automatically discover and invoke these skills

Omit --target to sync to the AboutSecurity repo itself (for using Agent directly in this repo). Re-run after adding or removing skills.

3. Using Dictionaries & Payloads

Dictionaries and payloads don't need syncing — just reference the paths in your Agent conversation:

"Use the dictionaries under /path/to/AboutSecurity/Dic/auth/ to brute-force SSH"
"Load the payload list from /path/to/AboutSecurity/Payload/xss/ for fuzz testing"

The Agent reads these files directly via Read / Glob tools — just provide the correct repo path.

Background (for newcomers): What is a Skill? Why sync?

• Skill = a structured methodology file (SKILL.md) that tells an AI Agent "what to do when encountering scenario X"
• Claude Code only recognizes the flat structure .claude/skills/<name>/SKILL.md
• This repo organizes skills in nested categories (e.g., skills/exploit/web-method/sql-injection/SKILL.md), the sync script creates symlinks for the nested → flat mapping
• After syncing, the Agent automatically matches and loads relevant Skills based on conversation context — no manual specification needed

4. Use via MCP Service (context1337) (Highly Recommended)

If you want to search and invoke all resources in this repo via natural language through AI assistants (Claude Code, Cursor, Claude Desktop, etc.), deploy context1337 — a standalone MCP resource service that turns AboutSecurity from a file repo into a consumable API (like context7, but for security).

git clone https://github.com/wgpsec/context1337.git
cd context1337
make run   # One command: clone data + build index + start server

Then add the MCP service to your AI tool:

# Claude Code
claude mcp add aboutsecurity --transport http http://localhost:8088/mcp

After that, query with natural language: "Search for SQL injection resources", "List all XSS payloads", "Find critical Apache vulnerabilities", etc. See context1337 README for details.

---

Skills Overview

skills/README.md covers the skills classification architecture, format specification, and benchmark testing process.

Dic/Payload Naming Conventions

Directory Naming

• All lowercase, hyphen-separated: file-backup/, api-param/, prompt-injection/

File Naming

• All English, lowercase, hyphen-separated
• No Fuzz_ prefix (legacy naming has been cleaned up)
• Examples: password-top100.txt, xss-tag-event-full.txt, complex-8char-upper-lower-digit.txt

_meta.yaml Metadata

Every directory containing data files has a _meta.yaml providing structured metadata for AI search:

category: auth                     # Top-level category
subcategory: password              # Subcategory (optional)
description: "Common weak passwords and complexity-rule generated password dictionaries"
tags: "password,weak-password,brute-force,login,credential"

files:
  - name: top100.txt
    lines: 100
    description: "Top 100 most common weak passwords"
    usage: "Initial brute-force screening, quick default password verification"
    tags: "top100,common,weak"

description and usage use Chinese, tags are bilingual (Chinese + English). Update the corresponding _meta.yaml when adding new dictionary/payload files.

Contributing

Read CONTRIBUTING.md before submitting, which covers Skill format specification, Vuln database writing standards, references requirements, and benchmark testing process.

References

• https://github.com/anthropics/skills/blob/main/skills/skill-creator/SKILL.md
• https://github.com/ljagiello/ctf-skills
• https://github.com/JDArmy/Evasion-SubAgents
• https://github.com/teamssix/twiki
• https://github.com/yaklang/hack-skills
• https://github.com/mukul975/Anthropic-Cybersecurity-Skills
• https://github.com/Pa55w0rd/secknowledge-skill
• https://github.com/0xShe/PHP-Code-Audit-Skill
• https://github.com/RuoJi6/java-audit-skills
• https://github.com/HackTricks-wiki/hacktricks
• https://github.com/HackTricks-wiki/hacktricks-cloud