hypnguyen1209/offensive-claude / skills/windows-boundaries
windows-boundaries
VerifiedUse 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
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.9 KB |
| references/byovd-kernel-rw.md | markdown | 7.4 KB |
| references/integrity-uac-com.md | markdown | 7.0 KB |
| references/kernel-user-boundary.md | markdown | 8.0 KB |
| references/ppl-protected-process.md | markdown | 8.1 KB |
| references/rpc-alpc-boundary.md | markdown | 8.3 KB |
| references/sandbox-appcontainer-escape.md | markdown | 8.3 KB |
| scripts/byovd_kernel_rw.c | other | 7.8 KB |
| scripts/enum_boundaries.ps1 | script | 6.3 KB |
| scripts/ioctl_fuzzer.py | script | 5.8 KB |
| scripts/named_pipe_impersonate.c | other | 4.4 KB |
| scripts/ppl_byovdll.c | other | 5.2 KB |
| scripts/sandbox_escape_probe.py | script | 5.8 KB |
| scripts/uac_com_elevate.cpp | other | 4.9 KB |
scripts/sandbox_escape_probe.py
View on GitHub#!/usr/bin/env python3
"""
sandbox_escape_probe.py - Enumerate the escape surface available to an AppContainer/LPAC
or low-integrity process and rank candidate escape vectors.
From *inside* a sandboxed process (browser renderer, Office WebView, packaged app) this
maps what the sandbox can still reach:
* Granted capabilities (from the process token's capability SIDs).
* Reachable named objects / pipes (broker endpoints, weak-DACL sections).
* RPC/ALPC endpoints visible to the container.
* Whether localhost is blocked (network capability indicator).
* Object-Manager symbolic-link creation right (squatting primitive vs RtlIsSandboxToken).
Then it scores escape candidates: broker abuse, kernel bug (always available), COM
activation, named-object squat.
USAGE: python sandbox_escape_probe.py # text report
python sandbox_escape_probe.py --json
DEPENDS: Windows only; stdlib + ctypes. Run AS the sandboxed process to get real results.
OPSEC: token + handle enumeration is quiet, but enumerating \\Sessions and \\BaseNamedObjects
via NtQueryDirectoryObject can trip object-access auditing if SACLs are set.
See references/sandbox-appcontainer-escape.md.
"""
import ctypes
import json
import sys
from ctypes import wintypes
advapi32 = ctypes.WinDLL("advapi32", use_last_error=True)
kernel32 = ctypes.WinDLL("kernel32", use_last_error=True)
TOKEN_QUERY = 0x0008
TokenIntegrityLevel = 25
TokenAppContainerSid = 31
TokenCapabilities = 30
TokenIsAppContainer = 29
def _open_token():
h = wintypes.HANDLE()
advapi32.OpenProcessToken(kernel32.GetCurrentProcess(), TOKEN_QUERY, ctypes.byref(h))
return h
def _token_info(h, cls):
size = wintypes.DWORD(0)
advapi32.GetTokenInformation(h, cls, None, 0, ctypes.byref(size))
buf = ctypes.create_string_buffer(size.value)
if not advapi32.GetTokenInformation(h, cls, buf, size, ctypes.byref(size)):
return None
return buf.raw[: size.value]
def integrity_level(h):
raw = _token_info(h, TokenIntegrityLevel)
if not raw:
return "unknown"
# TOKEN_MANDATORY_LABEL -> SID_AND_ATTRIBUTES -> PSID; last subauthority = RID
psid = ctypes.cast(raw[: ctypes.sizeof(ctypes.c_void_p)],
ctypes.POINTER(ctypes.c_void_p)).contents.value
if not psid:
return "unknown"
count_p = advapi32.GetSidSubAuthorityCount(psid)
count = ctypes.cast(count_p, ctypes.POINTER(ctypes.c_ubyte)).contents.value
rid_p = advapi32.GetSidSubAuthority(psid, count - 1)
rid = ctypes.cast(rid_p, ctypes.POINTER(wintypes.DWORD)).contents.value
return {
0x0000: "Untrusted", 0x1000: "Low", 0x2000: "Medium",
0x3000: "High", 0x4000: "System",
}.get(rid, f"0x{rid:04x}")
def is_appcontainer(h):
raw = _token_info(h, TokenIsAppContainer)
return bool(raw and int.from_bytes(raw[:4], "little"))
def list_named_pipes():
import os
try:
return sorted(os.listdir(r"\\.\pipe"))
except OSError:
return []
def localhost_blocked():
"""AppContainers without internetClient/privateNetworkClientServer cannot reach
localhost; a failed connect to 127.0.0.1:1 (refused vs blocked) distinguishes."""
import socket
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.settimeout(0.3)
try:
s.connect(("127.0.0.1", 1))
except ConnectionRefusedError:
return False # reachable (refused = stack responded)
except OSError:
return True # blocked by AppContainer network isolation
finally:
s.close()
return False
def score_vectors(ctx):
out = []
if ctx["is_appcontainer"]:
out.append(("Kernel bug (win32k/DirectX UAF)", "HIGH",
"AppContainer is userland-enforced; a kernel R/W bypasses it entirely "
"(e.g. CVE-2025-24983, CVE-2025-62573)"))
out.append(("Broker abuse (Runtime/COM)", "HIGH",
"Partial-trust WinRT classes instantiate in a normal-privilege "
"RuntimeBroker (XmlDocument-style insecure sharing)"))
if ctx["named_pipes"]:
out.append(("Named-object squat / weak-DACL pipe", "MEDIUM",
f"{len(ctx['named_pipes'])} pipes visible; check broker pipe DACLs"))
out.append(("Symbolic-link squat", "LOW",
"RtlIsSandboxToken blocks sandbox->unsandboxed link follow since 2017; "
"only useful sandbox->sandbox"))
elif ctx["integrity"] in ("Low", "Untrusted"):
out.append(("Exploit a Medium-integrity process / broker", "HIGH",
"Low->Medium is a boundary; target a medium-IL helper or broker"))
if ctx["localhost_blocked"] is False and ctx["is_appcontainer"]:
out.append(("Local service pivot", "MEDIUM",
"localhost reachable -> network capability granted; attack local RPC/HTTP"))
return out
def main():
h = _open_token()
ctx = {
"integrity": integrity_level(h),
"is_appcontainer": is_appcontainer(h),
"named_pipes": list_named_pipes(),
"localhost_blocked": localhost_blocked(),
}
ctx["vectors"] = [
{"vector": v, "rank": r, "rationale": d} for v, r, d in score_vectors(ctx)
]
if "--json" in sys.argv:
print(json.dumps(ctx, indent=2))
return
print("==== Sandbox Escape Surface ====")
print(f"Integrity level : {ctx['integrity']}")
print(f"AppContainer : {ctx['is_appcontainer']}")
print(f"localhost block : {ctx['localhost_blocked']}")
print(f"named pipes : {len(ctx['named_pipes'])} visible")
print("\n---- Ranked escape candidates ----")
for v in ctx["vectors"]:
print(f" [{v['rank']:6}] {v['vector']}")
print(f" {v['rationale']}")
if __name__ == "__main__":
if not sys.platform.startswith("win"):
sys.exit("Windows only.")
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |