Turn Claude into an Expert Penetration Tester with claude‑red: 58 Skills Across 12 Attack Domains

1. What It Is

claude‑red (author: 0xwilliamortiz) is a red‑team offensive security skill library designed for Claude’s skill system. Each attack area is represented by a structured SKILL.md file that contains the full methodology, tool list, attack paths, and escalation chains.

When a conversation mentions a keyword such as "SQL injection," Claude automatically loads the corresponding offensive‑sqli skill, instantly gaining expert‑level knowledge without needing additional explanation.

The library is intended for authorized red‑team penetration testing, bug‑bounty hunting, security research, CTF preparation, and training operators—use is at the user’s discretion.

2. Overview of the 12 Skill Categories

2.1 Web Applications (16 skills)

Largest section, covering OWASP Top 10 plus business‑logic flaws:

offensive‑sqli – error‑based, blind, OOB, DB specifics, ORM CVE, cloud paths

offensive‑xss – stored, reflected, DOM, Mutation XSS

offensive‑ssrf – server‑side request forgery, cloud metadata bypass

offensive‑ssti – server‑side template injection, engine detection, RCE paths

offensive‑xxe – XML external entity, OOB exfiltration, blind exploitation

offensive‑rce – remote code execution, command‑injection chains

offensive‑deserialization – Java/PHP/.NET gadget chains

offensive‑request‑smuggling – HTTP request smuggling, CL.TE/TE.CL, h2 desynchronisation

offensive‑waf‑bypass – WAF evasion, encoding tricks, case mutation

offensive‑graphql – GraphQL introspection bulk queries, IDOR via aliases

Additional items – IDOR, file upload, open redirect, parameter pollution, business‑logic bugs

2.2 Wireless Security (13 skills)

Professional section covering:

offensive‑wifi – 802.11 overview, wireless security basics

offensive‑wpa2‑psk – handshake capture, PMKID, hashcat 22000 password cracking

offensive‑wpa3‑sae – transition mode downgrade, Dragonblood, SAE side‑channel

offensive‑wpa‑enterprise – 802.1X/EAP attacks, eaphammer evil‑twin RADIUS

offensive‑wps – Pixie Dust offline attack, online PIN brute‑force, vendor PIN generators

offensive‑evil‑twin – KARMA/Mana attacks, captive portal, post‑association MITM

offensive‑krack‑fragattacks – KRACK + FragAttacks client testing

offensive‑deauth‑disassoc – directed/broadcast deauthentication, PMF awareness

offensive‑bluetooth‑ble – BLE GATT enumeration, pairing downgrade, sniffing, MITM

offensive‑bluetooth‑classic – BR/EDR classic Bluetooth, SPP/KNOB/HID spoofing

offensive‑zigbee‑thread‑matter – 802.15.4 mesh, Touchlink abuse, ZCL command injection

offensive‑z‑wave – Z‑Wave S0 key‑derivation flaw, S2 provisioning, hub pivots

offensive‑lorawan‑sub‑ghz – LoRaWAN ABP/OTAA, fixed‑code garage doors, TPMS

2.3 Active Directory (1 skill, expanding)

offensive‑active‑directory covers Kerberoasting, ASREProast, ACL abuse, ADCS ESC1‑15, delegation, persistence, hybrid AAD. The roadmap plans to split AD into 16 independent skills such as Kerberoasting, ASREProasting, ADCS, coercion, NTLM relay, BloodHound, ticket forgery, GPO abuse, etc.

2.4 Cloud Security (1 skill)

offensive‑cloud spans AWS, Azure, GCP – privilege escalation, IMDSv1 SSRF, cross‑account access, persistence, CSPM evasion. Cloud identity (Entra/AAD/Okta hybrid) skills are planned separately.

2.5 Other Categories

Mobile Security (1): Android + iOS – Frida, certificate pinning, storage, biometrics, deep links

IoT Security (1): hardware reconnaissance, firmware, RTOS, ICS/OT, MQTT/CoAP – to be split into 10 skills

Infrastructure & Red Team (7): initial access, EDR evasion, shellcode, keyloggers, Windows mitigation bypass

Vulnerability Development (6): stack/heap exploits, ROP chains, modern kernel mitigations (ASLR/CFG/CET/PAC), TOCTOU

Fuzzing & Vulnerability Research (4): libFuzzer, AFL++, coverage‑guided, vulnerability taxonomy

Reconnaissance (2): OSINT tools (recon‑ng, theHarvester, Maltego) and methodology

AI Security (1): prompt injection, jailbreak, RAG poisoning

Practical Tools (2): rapid vulnerability triage checklist, professional penetration‑test report writing

3. Quick Deployment

claude‑red requires a Node.js environment.

git clone https://github.com/0xwilliamortiz/claude-red ~/.claude/skills/claude-red
cd ~/.claude/skills/claude-red
./install.sh

Or install via npm:

npm install -g .
claude-red

After installation, Claude automatically loads the appropriate skill when a trigger phrase like "SQL injection" is mentioned.

You can also load skills manually:

# List all skills
claude-red list

# Install only a category
./install.sh --category web

# View skills in a category
claude-red --category web

4. Technical Highlights

Conversational Triggering: No manual context switching; Claude decides which skill to load based on the question, saving tokens.

SKILL.md Standardization: Each skill is a structured markdown file containing methodology, tools, and trigger‑word explanations, which can be extracted as system prompts.

Sparse Checkout Support: Using Git sparse‑checkout, you can pull only the needed categories, reducing time and storage.

Coverage of Emerging Areas: Includes AI security (prompt injection, jailbreak, RAG poisoning) and wireless security (BLE, Zigbee, Z‑Wave, LoRa), which are often weak points in traditional penetration‑testing toolchains.

5. Safe‑Use Guidelines

claude‑red is a pure methodology knowledge base; the install.sh script only copies SKILL.md files locally and contains no malicious code.

However, recent community discussions have raised security warnings about the project. Users should review the latest repository commits for anomalies and run the tool in an isolated environment, avoiding direct use against live targets.

6. Project Information

Open‑source URL: https://github.com/0xwilliamortiz/claude-red

Number of Skills: 58 (continuously expanding)

Number of Categories: 12

Applicable Scenarios: authorized penetration testing, bug‑bounty hunting, security research, CTF

Copyright Notice: This article is originally published by 华盟网 and all rights are reserved.