Windows Boundaries logo

Windows Boundaries

Community
hypnguyen1209
windows-boundaries

Use when crossing a Windows security boundary or escaping a sandbox — kernel/user crossing (win32k/dxgkrnl UAF CVE-2025-24983), BYOVD kernel R/W, UAC/COM elevation, AppContainer/LPAC & Chromium-Mojo sandbox escape (CVE-2025-2783), PPL bypass, RPC/ALPC & named-pipe impersonation

Overview

Publisherhypnguyen1209
Repositoryoffensive-claude
Skill namewindows-boundaries
Stars
396
Forks
71
Bundled files
13
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 13 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by hypnguyen1209 on GitHub. Read the source before you install it.

Installation

Install the Windows Boundaries AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/hypnguyen1209/offensive-claude/tree/main/skills/windows-boundaries
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/hypnguyen1209/offensive-claude.git /tmp/offensive-claude
mkdir -p .claude/skills
cp -r /tmp/offensive-claude/skills/windows-boundaries .claude/skills/windows-boundaries
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Windows Boundaries in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Windows Boundaries on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Windows Boundaries is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Windows Security Boundaries

When to Activate

  • Planning a privilege-escalation path that crosses a Windows security boundary (integrity level, AppContainer/LPAC, PPL, or the kernel/user line)
  • Sandbox-escape research: browser renderer (Chromium/Edge Mojo), Office WebView, packaged apps, AppContainer/LPAC brokers
  • Reaching Ring 0 via win32k/dxgkrnl bugs or BYOVD for a kernel read/write primitive
  • Defeating PPL to dump LSASS or tamper with EDR self-defense
  • Going from a SeImpersonate service account to SYSTEM via RPC/ALPC/named-pipe abuse
  • UAC bypass (Medium → High) via auto-elevating COM or registry hijack

Boundary stack (high → low): VTL1 (Secure Kernel/Cred Guard) > Ring 0 (ntoskrnl/win32k/drivers) > Ring 3: System > High > Medium > Low > AppContainer/LPAC. PPL is an orthogonal wall guarding LSASS/EDR even from SYSTEM.

Technique Map

TechniqueATT&CKCWEReferenceScript
win32k / dxgkrnl UAF → kernel R/W → token stealT1068CWE-416references/kernel-user-boundary.mdscripts/ioctl_fuzzer.py
Driver IOCTL abuse (METHOD_NEITHER arb-R/W)T1068CWE-781references/kernel-user-boundary.mdscripts/ioctl_fuzzer.py
BYOVD load + MSR/LSTAR or phys R/W → SYSTEMT1068, T1543.003CWE-1188references/byovd-kernel-rw.mdscripts/byovd_kernel_rw.c
BYOVD EDR-kill (callback nulling)T1562.001, T1014CWE-1188references/byovd-kernel-rw.mdscripts/byovd_kernel_rw.c
UAC bypass — ICMLuaUtil elevated COM monikerT1548.002CWE-269references/integrity-uac-com.mdscripts/uac_com_elevate.cpp
UAC bypass — fodhelper HKCU registry hijackT1548.002, T1112CWE-269references/integrity-uac-com.mdscripts/uac_com_elevate.cpp
AppContainer/LPAC broker abuse / cap over-grantT1211CWE-668references/sandbox-appcontainer-escape.mdscripts/sandbox_escape_probe.py
Chromium/Edge Mojo IPC sandbox escapeT1211CWE-501references/sandbox-appcontainer-escape.mdscripts/sandbox_escape_probe.py
Named-object / symbolic-link squattingT1211CWE-59references/sandbox-appcontainer-escape.mdscripts/sandbox_escape_probe.py
PPL bypass — BYOVDLL (old signed DLL into PPL)T1003.001, T1211CWE-426references/ppl-protected-process.mdscripts/ppl_byovdll.c
PPL bypass — live-dump / WER / Protection-clearT1003.001, T1562.001CWE-269references/ppl-protected-process.mdscripts/byovd_kernel_rw.c
RPC server spoof / endpoint squat (PhantomRPC)T1559, T1134.001CWE-287references/rpc-alpc-boundary.mdscripts/named_pipe_impersonate.c
Named-pipe client impersonation (Potato family)T1134.002, T1134.001CWE-294references/rpc-alpc-boundary.mdscripts/named_pipe_impersonate.c
Host boundary posture enumerationT1082CWE-693(all)scripts/enum_boundaries.ps1

Quick Start

cmd
:: 0. Map the host's boundary posture -> pick the right primitive
powershell -ep bypass -File scripts/enum_boundaries.ps1
::    reports: integrity, AppContainer, SeImpersonate/SeDebug, HVCI, driver blocklist,
::    LSASS RunAsPPL, drivers loaded from user-writable paths, named-pipe count.

:: 1. From a sandbox (renderer/packaged app): rank escape vectors
python scripts/sandbox_escape_probe.py

:: 2. Medium -> High: UAC bypass (no file dropped via COM moniker)
uac_com_elevate.exe com "C:\Windows\System32\cmd.exe /c whoami /groups > C:\poc.txt"

:: 3. SeImpersonate present -> SYSTEM via PhantomRPC endpoint squat
whoami /priv | findstr SeImpersonate
named_pipe_impersonate.exe \\.\pipe\W32TIME "C:\Windows\System32\cmd.exe"
w32tm /resync                          :: triggers the SYSTEM client to connect

:: 4. Need Ring 0: map a driver's IOCTLs, then BYOVD kernel R/W -> token steal
python scripts/ioctl_fuzzer.py --device RTCore64 --map
byovd_kernel_rw.exe C:\test\driver.sys token

:: 5. Dump LSASS under PPL without a driver (BYOVDLL / CNG KSP)
ppl_byovdll.exe register C:\stage\vuln_ncryptprov.dll GhostKsp
ppl_byovdll.exe trigger  GhostKsp

OPSEC & Detection (summary)

TechniqueTelemetry / IOCDetection (Sigma / EDR)OPSEC note
win32k/dxgkrnl exploitBugcheck EID 1001 ref win32k/dxgkrnl; process token flips to S-1-5-18 with no service genealogySigma on bugchecks + token anomaly; ETW Microsoft-Windows-Win32kPrefer data-only token copy over CFG/CET-fighting control-flow hijack; missed race = loud bugcheck
BYOVDSysmon EID 6 driver load from user-writable path; 7045/4697 kernel service; EDR telemetry goes silentSigma on unusual driver path + known-bad hashes (loldrivers.io); driver-load-time windowUse un-blocklisted/HVCI-compatible driver; restore LSTAR/callbacks; DeleteService + del .sys
UAC bypass (COM/fodhelper)Sysmon EID 13 write to HKCU ms-settings\Shell\Open\command; dllhost /Processid:{3E5FC7F9..} spawning shellElastic/ManageEngine ICMLuaUtil + fodhelper rulesRotate to a less-known AutoElevate CLSID; don't spawn cmd from dllhost; clean HKCU tree
Sandbox escapeRenderer/GPU process spawning cmd/powershell/rundll32; Mojo handle-transfer anomaliesSigma on browser child-process; patch-posture alert (Edge<134.0.3124.93)Stay in-process post-escape (IPC layer is an EDR blind spot); symlink squat is mostly dead post-2017
PPL bypassOld keyiso/ncryptprov version loaded in lsass (EID 7); EID 10 lsass read; Cryptography\Providers write; live-dumpSigma on DLL version-mismatch + GrantedAccess to lsassDriverless (BYOVDLL/WER/live-dump) avoids BYOVD noise; restore Protection byte; obfuscate dumps
RPC/ALPC impersonation4624 logon type 9 (Advapi); SYSTEM child of service-acct parent; RPC_S_SERVER_UNAVAILABLE (EID 1)Elastic named-pipe impersonation rule; ETW RPC + high impersonation levelPhantomRPC endpoint squat blends w/ admin triggers (w32tm/gpupdate); RevertToSelf + close pipe

Deep Dives

  • references/kernel-user-boundary.md — Ring 3→Ring 0 via win32k/dxgkrnl UAF (CVE-2025-24983, CVE-2025-62573, CVE-2025-55224, CVE-2025-62221, CVE-2026-26132), driver IOCTL surface, kernel R/W → EPROCESS token steal / Protection clear / PreviousMode, modern no-PTE-self-map constraint, KDP/HVCI considerations.
  • references/byovd-kernel-rw.md — Bring-Your-Own-Vulnerable-Driver: load signed driver, MSR-LSTAR vs physical-memory primitives (CVE-2025-8061), EDRKillShifter / Qilin driver rotation, 2025-2026 driver+CVE catalog, blocklist/HVCI bypass, driver-load detection.
  • references/integrity-uac-com.md — integrity model, appinfo/autoElevate internals, fodhelper (UACME 33) + ICMLuaUtil/CMSTPLUA (UACME 41), 2025 IEditionUpgradeManager bypass, finding new AutoElevate COM classes, T1548.002 ITW campaigns.
  • references/sandbox-appcontainer-escape.md — AppContainer/LPAC model + capabilities, broker abuse (WinRT XmlDocument), Chromium/Edge Mojo escapes (CVE-2025-2783 ForumTroll, CVE-2025-4609 ipcz), symbolic-link/RtlIsSandboxToken mitigation, 3-bug renderer chain.
  • references/ppl-protected-process.md — PS_PROTECTION/signer levels, BYOVDLL "Ghost in the PPL" (CVE-2023-28229/36906, BCryptRegisterProvider), NtSystemDebugControl live-dump (Win11 23H2), WER dump, kernel Protection-clear, PPLBlade, Credential Guard hard stop.
  • references/rpc-alpc-boundary.md — named-pipe client impersonation primitive, ALPC SQoS/RequiredServerSid, PhantomRPC server-spoof/endpoint-squat (2026, unpatched), Potato family (PrintSpoofer/GodPotato/SigmaPotato), WER ALPC LPE, NtObjectManager enumeration.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Windows Boundaries AI skill do?

Use when crossing a Windows security boundary or escaping a sandbox — kernel/user crossing (win32k/dxgkrnl UAF CVE-2025-24983), BYOVD kernel R/W, UAC/COM elevation, AppContainer/LPAC & Chromium-Mojo sandbox escape (CVE-2025-2783), PPL bypass, RPC/ALPC & named-pipe impersonation

Why use Windows Boundaries on TypingMind?

Because you install it once and use it with any model. Windows Boundaries is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Windows Boundaries in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/hypnguyen1209/offensive-claude/tree/main/skills/windows-boundaries. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Windows Boundaries?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Windows Boundaries?

As many as you like. As long as a model supports skills, you can use Windows Boundaries with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Windows Boundaries AI skill free?

Yes. It is published on GitHub by hypnguyen1209 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇