How to use the Penetration Testing Toolchain agent skill (pentest-tools)

What the skill makes the agent do

An agent skill is a set of instructions the AI loads when a request matches its description. This one covers reconnaissance (port scanning, subdomain enumeration, service identification), vulnerability scanning, web testing such as SQL injection and directory brute-forcing, and password cracking.

It maps common tools to the commands to use, for example Nmap for port and service scanning, Nuclei for template-based vulnerability scanning, SQLMap for SQL injection and FFUF for directory and parameter fuzzing. It points the agent to an MCP server (pentestMCP or mcp-security-hub) that exposes those tools to the agent.

• Reconnaissance: Nmap, Masscan, Subfinder, httpx
• Vulnerability scanning: Nuclei, ZAP, Nikto
• Web testing: SQLMap, FFUF, Gobuster, XSStrike
• Password cracking: Hashcat, John the Ripper, Hydra

Install the skill (ZIP download or Claude apps)

Download the skill ZIP from this page and unzip it. For Claude Code, place the unzipped folder in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project only.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills

Set up an MCP backend

The skill supports two MCP backends and you only need one. The first option, pentestMCP, packages 20+ tools into a single Docker container, so it requires Docker. The second, mcp-security-hub, provides each tool as a separate MCP server that you enable as needed. You can also register a single-tool MCP server for one tool.

• pentestMCP (Docker, all-in-one): exposes Nmap, Nuclei, ZAP, SQLMap, FFUF, Nikto, Gobuster, Subfinder and httpx
• mcp-security-hub (modular): separate servers for Nmap, Ghidra, Nuclei, SQLMap and Hashcat; follow each sub-module's README
• Single-tool option: nmap-mcp-server, nuclei-mcp, or the SQLMap module in mcp-security-hub

Trigger the skill with natural requests

You do not need a special command. Describe the task in plain language and include the target and the kind of work. The skill's description lists trigger terms such as 渗透测试 (penetration test), 端口扫描 (port scan), Nmap, 漏洞扫描 (vulnerability scan), Nuclei, SQL 注入 (SQL injection), SQLMap, 目录爆破 (directory brute-force), FFUF, 密码破解 (password cracking), Hashcat, 子域名 (subdomain), ZAP and Burp.

Before any scan, the skill requires a scope file (scope.md) with authorization granted. Mention it in your request so the agent can find it.

• "Using the scope in scope.md, run an Nmap service scan on 203.0.113.10 and summarise open ports."
• "Run Nuclei with critical and high severity templates against the staging site listed in scope.md."
• "Enumerate subdomains of example.com in scope, then check which ones respond over HTTP."

Example workflows

The skill describes a standard flow: reconnaissance first, then vulnerability scanning, then exploitation only where authorized, then reporting. It also includes a five-minute quick scan: an Nmap service scan, a Nuclei scan limited to critical and high severity, a directory check with FFUF if a web service is present, then a summary to decide next steps.

Authorized work is logged step by step: each command and its result should be recorded so it can feed a report.

• Quick scan: nmap -sV -sC target, then nuclei -u target -severity critical,high
• Directory discovery: ffuf -u target/FUZZ -w common.txt
• SQL injection check: sqlmap -u "url?id=1" --batch --dbs

What you need before you start

You need either Docker (for pentestMCP) or the individual tools installed yourself, such as Nmap, Nuclei, SQLMap and FFUF. You also need an MCP-capable agent client to register the MCP server, and a written scope or authorization file that the workflow references.

Reqable and BurpSuite are optional, for the related MCP integrations. The source does not list any other API keys or accounts beyond what is in the SKILL.md file, so check that file if a feature asks for one.

• Docker, or Nmap, Nuclei, SQLMap, FFUF installed locally
• An MCP-capable agent client
• A written scope/authorization file
• Optional: Reqable desktop client, BurpSuite

Safety: scripts, network and authorization

This skill runs scripts and uses the network. Its scans and exploitation steps can be illegal and harmful against systems you do not own or have explicit written permission to test. Confirm the target is in scope before every scan.

Control scan rate to avoid triggering WAF or IDS alerts or crashing targets, and do not run scans that could disrupt production. The skill says the AI should wait for confirmation at each critical step rather than automate blindly.

• Only test systems you own or have written permission to test
• Keep scan rates controlled
• Prefer passive recon before active scanning
• Do not run scans against production without a plan
• Keep a log of every command and result

Tips for getting better results

Give the agent a precise target and scope, and name the tools you want it to use. Ask for a summary after each batch of results so you can confirm the next step. If a tool is missing, the skill is designed to check availability and point you to installation steps rather than guess paths.

For Reqable MCP, the recommended default scope is minimal; the all scope exposes tools that can change proxies, rules or saved data. Captured traffic and replay still need to fall within your scope file.

• Name the scope file and target explicitly
• Start with low-impact tools like Nmap service detection
• Ask the agent to log each step
• Use minimal scope for Reqable MCP