Professional Autonomous Pentesting

AI-WAR
CYBER PLATFORM

AI 战争 · 网络平台

ai-war agent platform (war.xcom.dev)

AI-driven autonomous pentesting infrastructure for cybersecurity professionals. Orchestrate 20+ integrated tools through a single LLM-powered attack chain. For authorized network testing only. OSINT → Enumeration → Exploitation → Reporting.

Scroll
00

Download Attack Chain

下载攻击链

Deploy the autonomous pentesting orchestrator on your infrastructure. The attack chain script handles tool orchestration, authorization checking, and multi-stage scan execution. Full scope verification required before scan initiation.

Quick Deploy (Curl)

$ bash deployment
curl -O https://war.xcom.dev/attack-chain.sh
bash attack-chain.sh --target example.com --mode standard

✓ Requires: nmap, nikto, gobuster (optional: enhanced results)
✓ Output: JSON/Markdown/PDF report
✓ Authorized scope verification required
✓ Scans: OSINT → Enumeration → Web → Exploitation

Advanced Usage

./attack-chain.sh \
  --target TARGET_DOMAIN \
  --mode [quick|standard|thorough|continuous] \
  --output /path/to/output/ \
  --format [json|markdown|html|sarif] \
  --scope-token YOUR_AUTH_TOKEN

✓ Full parametrization for CI/CD integration
✓ Docker deployment: docker run war.xcom.dev:latest
✓ Kubernetes manifests available
✓ See: ./attack-chain.sh --help

00.5

Interactive Scanning Interface

交互式扫描界面

Download the cross-platform war.xcom.dev GUI client for macOS, Linux, and Windows. Point-and-click target selection, real-time scan progress, and one-click report export.

Scan Interface

Features

  • Real-time Progress
    Live scan stage updates. Cancel anytime.
  • Multi-Format Export
    JSON, Markdown, PDF, HTML, SARIF output formats.
  • CVSS Scoring
    Every finding auto-scored. CRITICAL/HIGH/MEDIUM/LOW filtering.
  • Local Execution
    Runs on your infrastructure. No cloud scanning. Full privacy.
  • History & Comparison
    Track scan history. Compare findings over time. Detect drift.
Download GUI (v2.1.0)
macOS (Intel/Apple Silicon) • Linux • Windows • Docker
00.7

AI Agent Orchestration

AI 智能体编排

war.xcom.dev is fully integrated with the OpenClaw AI agent framework. Manage, customize, and orchestrate pentesting agents from a unified workspace. Chain multiple skills together. Deploy autonomous red teams.

🎯

Terminator Agent

Primary autonomous pentesting agent. Orchestrates the full attack chain. Handles scope verification, tool sequencing, and report synthesis.

Edit in OpenClaw →
🛠

Pentest Skill (D2)

Full D2 specification for pentesting workloads. SKILL.md + TOOLS.md + REPORT.md. Integrated tool chain definitions. Customizable scan templates.

View Specification →
🔍

Hunter Skill

Continuous vulnerability hunting and asset monitoring. Real-time alerts. Feeds into pentesting pipeline. Detects infrastructure drift and new exposures.

Configure Monitoring →

Access the OpenClaw agent management interface at ai.xcom.dev/workspace. Create custom agents, chain skills, set trigger conditions, and deploy autonomous red teams to your network.

Open OpenClaw Workspace
01

Full-Stack Attack Surface Coverage

全栈攻击面覆盖

war.xcom.dev is the autonomous pentesting surface of the XCOM network. A single command triggers a multi-stage scan pipeline — OSINT first, then network enumeration, web application probing, credential testing, and LLM-powered synthesis of the full attack surface. Every finding is CVSS-scored, de-duplicated, and exported in your format of choice. For authorized security professionals only.

20+
Integrated Tools
4
Scan Modes
6
Report Formats
LLM
Attack Synthesis
02

The Pentest Arsenal

渗透测试工具库

Every tool in the stack is open source, battle-tested, and orchestrated via the XCOM pentest skill. Click any tool to view its upstream documentation.

OSINT & Reconnaissance

Network & Infrastructure

Web Application

Credentials & Exploitation

03

Four Levels of Depth

四层深度扫描

Each mode progressively deepens coverage and increases dwell time. Use Quick for rapid triage, Standard for typical engagements, Thorough for comprehensive red teams, and Continuous for persistent monitoring loops.

Mode Duration Tools Active Use Case
Quick 5–15 min Nmap top-1000, Nikto, Gobuster common, Shodan passive Rapid triage. First look at unknown asset. CTF reconnaissance.
Standard 30–90 min Full TCP Nmap, Gobuster + FFUF, Nikto, sqlmap detect, Amass, ZAP passive Standard pentest engagement. VAPT scoping phase. Bug bounty initial recon.
Thorough 2–6 hrs All tools. UDP scan, ZAP active, sqlmap exploit, Hydra credential test, Metasploit CVE validation, TruffleHog git scan Full red team engagement. Compliance audit. Comprehensive VAPT report.
Continuous
24/7
Persistent Lightweight diff-based. Amass + Shodan polling. New port/service delta alerts. Attack surface monitoring. Detect new exposures. Notify on configuration drift.
04

LLM-Orchestrated Pipeline

LLM 驱动的流水线

Every scan follows the same structured attack chain. Each stage feeds its findings into the next. The LLM synthesis layer correlates all outputs into a unified attack narrative with prioritised remediation.

1

Scope Gate

Authorization check. Verify written scope. Apply production blocklist. Log operator identity and scan parameters to immutable audit trail.

2

OSINT Layer

Amass subdomain enum → theHarvester email/IP harvest → crt.sh cert scrape → Shodan/Censys/VT/OTX passive intel → DNS zone transfer attempts.

3

Network Enumeration

Nmap SYN stealth full TCP + UDP top-200 → service version detection → NSE vuln scripts → smbmap + CrackMapExec on Windows targets → SNMP enumeration.

4

Web Application

Nikto banner + vuln scan → Gobuster dir + DNS → FFUF parameter fuzzing → ZAP spider + active scan → GraphQL introspection → JWT decode → CORS check.

5

Database & Auth

sqlmap on discovered injection points (w/ tamper scripts for WAF bypass) → NoSQL injection probes → Hydra/Medusa credential testing → IDOR enumeration.

6

LLM Synthesis

All findings fed to local Qwen model. Correlates CVEs, chains attack vectors, estimates impact, assigns CVSS scores, and generates human-readable attack narrative.

ORCHESTRATION SCRIPT

# xcom-pentest standard mode TARGET="example.com" MODE="standard" OUTPUT="/tmp/xcom-pentest-$(date +%s)" # Stage 1 — OSINT (parallel) amass enum -d $TARGET & theHarvester -d $TARGET -l 300 -b all & curl -s "https://crt.sh/?q=$TARGET&output=json" & wait # Stage 2 — Network nmap -sS -sV -O --script vuln \ -p- --min-rate 2000 $TARGET # Stage 3 — Web nikto -host $TARGET -output $OUTPUT/nikto.txt gobuster dir -u https://$TARGET \ -w /usr/share/seclists/Discovery/\ Web-Content/common.txt ffuf -u https://$TARGET/FUZZ \ -w /usr/share/seclists/Fuzzing/\ fuzz-Bo0oM.txt -mc 200,301,302,403 # Stage 4 — LLM synthesis xcom-analyze --findings $OUTPUT/ \ --model qwen3.5:35b \ --output $OUTPUT/report.md
05

Six Export Formats

六种导出格式

Every scan auto-generates a full report. CVSS v3.1 scoring on every finding. De-duplicated across tool outputs. Remediation steps linked to CVE references.

MD

Markdown

Human-readable report with executive summary, per-host findings, CVSS table, and remediation checklist. Renders natively in GitHub, Obsidian, and Notion.

{}

JSON

Machine-readable structured output. Feeds directly into SIEM pipelines, ticketing systems (Jira, Linear), and CI/CD security gates.

HTML

HTML / PDF

Branded client-ready report. Full CVSS risk matrix, executive summary, and technical appendix. Converts to PDF via headless Chrome.

SA

SARIF

Static Analysis Results Interchange Format. GitHub Code Scanning compatible. Surfaces findings directly in pull request annotations and security dashboard.

Telegram / WhatsApp

Live scan progress and summary alerts delivered to your channel. Critical findings (CVSS ≥ 9.0) trigger immediate push notifications.

GH

GitHub Issues

Auto-creates labelled issues from findings. CRITICAL/HIGH/MEDIUM/LOW severity labels, CVSS score in title, full technical detail and remediation in body.

06

Quick Start

快速开始

The XCOM pentest skill integrates with the OpenClaw agent framework. Trigger scans via Telegram, WhatsApp, or direct CLI. All commands require a valid scope authorization token.

TELEGRAM / WHATSAPP

Message your XCOM agent from any device. The skill handles scope verification, orchestrates the scan, and pushes progress updates every 5 minutes.

# Start a scan /pentest scan example.com --mode standard # Get latest report /pentest report --format markdown # Compare two scans /pentest compare scan-001 scan-002 # Check scope /pentest scope example.com

CLI / OPENCLAW SKILL

Run the pentest skill directly from the OpenClaw agent framework. The full D2 skill specification is available at the link below.

# Install skill dependencies sudo apt install nmap nikto gobuster ffuf \ sqlmap amass hydra zaproxy # Run via OpenClaw openclaw skill run pentest \ --target example.com \ --mode thorough \ --output /tmp/report/ # Skill specification ~/.openclaw/workspace/skills/pentest/ ├── SKILL.md # Full D2 spec ├── TOOLS.md # Install guide └── REPORT.md # Report templates
View Skill Protocol

Start Your First Scan

Authorize a target. Trigger the chain. Receive a full attack surface report in under 90 minutes.