Agent Skills

ntobjmanager-mcp

A Model Context Protocol (MCP) server that gives AI agents live, stateful Windows RPC attack-surface research, built on James Forshaw's NtObjectManager (NtCoreLib).

README.md

πŸ›°οΈ NtObjectManager-MCP

Stateful Windows RPC Research MCP β€” 2024–2026 CVE methodologies as one-click tools

Python License PowerShell MCP

🌐 δΈ­ζ–‡η‰ˆ


What is NtObjectManager-MCP?

A Model Context Protocol server that gives an AI agent live, stateful access to Windows RPC attack-surface research, built on James Forshaw's NtObjectManager (NtCoreLib).

Three things a generic PowerShell MCP cannot do β€” and the reason this exists:

  1. Stateful RPC connections β€” a persistent PowerShell engine keeps parsed RpcServer objects and connected RPC clients alive across tool calls: rpc_connect once, rpc_call many times (auth handshakes, context-handle chains, session variables survive).
  2. CVE methodology as fixed tools β€” the standard hunting workflows from 2024–2026 public research are one-click, not prompt-engineering.
  3. Stateful execution inside a lab VM β€” the same one-engine principle applied in the guest: rpc_vm_exec keeps variables and connected RPC clients alive across calls through a single persistent guest runspace, never a fresh shell per call (the vmrun fallback is reported as stateful: false).
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  AI Agent (Claude Code / OpenCode / any MCP client)                β”‚
β”‚      β”‚  MCP (stdio, 24 tools)                                      β”‚
β”‚      β–Ό                                                             β”‚
β”‚  server.py ── snippets.py (PS templates, @@TOKEN@@ + ps_str escape)β”‚
β”‚      β”‚                                                             β”‚
β”‚      β–Ό                                                             β”‚
β”‚  ps_engine.py ── persistent powershell.exe (base64 + __MCP_DONE__) β”‚
β”‚      β”‚            state: $RPCMCP = @{ Servers; Clients; vars }     β”‚
β”‚      β–Ό                                                             β”‚
β”‚  NtObjectManager / NtCoreLib  ──►  RPC runtime (ALPC / pipe / TCP) β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Tool Matrix (24)

Core stateful pipeline

Tool Purpose
rpc_parse(file, symbol_path?) Parse a PE for RPC servers, cache (keys file_N)
rpc_state() Cached servers + live sessions
rpc_get_interface(key) Procedures, NDR params, context handles, strictness
rpc_query_endpoints(ifid?, search_binding?, find_alpc_port?) EPM query (local or remote)
rpc_running_servers(pid?/service?) Live process/service enumeration
rpc_connect(session, key, binding?, auth?) Generate + connect client (stateful)
rpc_methods(session) Signatures with opnum mapping
rpc_call(session, method, args_json, store_as?) Reflection invoke; {"__var__"} passes stored objects
rpc_disconnect(session) Drop session

VM lab bridge (stateful guest execution)

Tool Purpose
rpc_vm_exec(ps, timeout?, vm?) Run PowerShell inside a lab VM; state survives across calls (persistent guest runspace)
rpc_vm_start_listener(vm?) Deploy/start the persistent guest HTTP engine (vm_listener.ps1)

2024–2026 CVE methodology tools

Tool Methodology source
rpc_scan_context_handles(paths) Context-handle type confusion β€” CVE-2025-48815 pattern (whereisk0shl 2026)
rpc_inventory(paths?, limit?) Attack-surface inventory + EPM cross-check β€” MS-RPC-Fuzzer phase 1 (CVE-2025-26651)
rpc_fuzz(session, dry_run=True) Primitive-only default-value fuzzing with ok/denied/error classification β€” dry-run by default
rpc_find_hijackable() Unregistered interfaces of stopped services β€” EPM poisoning / RPC-Racer (CVE-2025-49760/59200/59230)
rpc_etw_unreachable(duration, trigger_script?) Clients calling dead servers β€” PhantomRPC (Kaspersky 2026), admin required
rpc_interface_security(key) ALPC SD / anonymous-ACE audit β€” MS-NRPC null session (SafeBreach/Securelist 2025)
rpc_decode_flags(flags) RpcServerRegisterIf3 flag bitmask decoding
rpc_format_client(key) Export generated C# client source (offline grep workflow)
rpc_new_struct(session, type, store_as) Build NDR complex types as session vars
rpc_alpc_squat(name, duration) ALPC port squat + connection capture (race validation primitive)
rpc_accessible_tasks() User-startable tasks (Dark-Elevator chain material, CVE-2026-66804 pattern)
rpc_vars / rpc_clear_cache Session-variable and cache management (eviction cap 150)

Every tool call is appended to output/mcp_audit.log.

Field-Tested (real machine, full hunting round)

Candidate Result
srvsvc.dll 98716d03… flagged HIGH Identified as XactSrv (XsOpenPrinter/XsClosePrinter/XsAddJob/XsScheduleJob) β€” single printer-handle type; live probe: non-admin connect OK but XsOpenPrinter β†’ ACCESS_DENIED (authorization gate works). Scanner false-positive mode documented
ssdpsrv.dll (CVE-2025-48815 original) All 20 context handles strict β€” patched state on current builds
51-module sweep 31 findings, 6 HIGH, all "one producer β†’ many consumers"; NDR layer cannot prove multi-type handles (needs RE)
EPM hijack surface 10 stopped services with unregistered interfaces (AppIDSvc, ClipSVC, dcsvc…)
Task chains 44 user-startable SYSTEM tasks inventoried
Verdict No confirmable exploitable vuln on the tested host β€” with per-step evidence

πŸš€ Quick Start

# 1) Prerequisites (one-time)
Install-Module NtObjectManager -Scope CurrentUser -Force
pip install -r requirements.txt            # mcp>=1.2.0 (1.x / 2.x compatible)

# 2) Verify β€” three suites, all green
python tests\smoke_test.py                 # 17 checks (live MCP stdio round-trip)
python tests\var_test.py                   # 10 checks (store_as / __var__ mechanics)
python tests\audit.py                      # 43 checks (edge cases, hostile paths, concurrency)

# 3) Run the server
python server.py                           # stdio MCP

Claude Code:

claude mcp add ntobjectmanager-rpc -- python C:\path\to\ntobjmanager-mcp\server.py

Any MCP client (e.g. OpenCode opencode.json):

{
  "mcp": {
    "ntobjectmanager-rpc": {
      "type": "local",
      "command": ["python", "C:\\path\\to\\ntobjmanager-mcp\\server.py"],
      "enabled": true
    }
  }
}

Example: context-handle type confusion (CVE-2025-48815 pattern)

1. rpc_parse  C:\Windows\System32\target.dll  [symbol_path optional]
2. rpc_scan_context_handles ["C:\\Windows\\System32\\target.dll"]
3. rpc_get_interface target_0                 β†’ producer ([out] ctx) / consumer ([in] ctx) pairs
4. rpc_connect s1 target_0                    β†’ auto-discovers binding via EPM
5. rpc_methods s1                            β†’ opnum-mapped signatures
6. rpc_call s1 XsOpenPrinter-like args store_as="h"   β†’ keep the raw handle object
7. rpc_call s1 XsClosePrinter-like [{"__var__":"h"}] β†’ feed it to the other type

store_as / {"__var__"} is the core chain primitive: RPC return objects flow between calls without serialization round-trips, which is exactly what producer→consumer handle-confusion testing needs.

πŸ“ Project Structure

ntobjmanager-mcp/
β”œβ”€β”€ server.py            # 24 MCP tools + audit logging wrapper
β”œβ”€β”€ snippets.py          # PowerShell templates (@@TOKEN@@ render + ps_str escaping)
β”œβ”€β”€ ps_engine.py         # Persistent engine: base64 cmds + __MCP_DONE__ markers, timeouts
β”œβ”€β”€ wrapper.ps1          # PS-side loop (state lives in $RPCMCP)
β”œβ”€β”€ vm_listener.ps1      # Persistent guest HTTP bridge (stateful VM exec)
β”œβ”€β”€ tests/
β”‚   β”œβ”€β”€ smoke_test.py    # 17 checks β€” live stdio end-to-end
β”‚   β”œβ”€β”€ var_test.py      # 10 checks β€” store_as/__var__ object passing
β”‚   β”œβ”€β”€ audit.py         # 43 checks β€” hostile inputs, concurrency, engine kill/restart
β”‚   β”œβ”€β”€ hunt.py          # Full dogfood hunting round (safe policy)
β”‚   β”œβ”€β”€ hunt2_static.py  # Deep-dive: symbols + producer/consumer map
β”‚   β”œβ”€β”€ hunt2_wide.py    # 51-module sweep
β”‚   └── hunt2_probe.py   # Safe runtime probes (exposure / task cross-ref)
β”œβ”€β”€ ARCHITECTURE.md      # Engine protocol + design decisions
β”œβ”€β”€ CHANGELOG.md         # Decision history (R1–R13)
β”œβ”€β”€ SECURITY.md          # Authorized-use policy + MSRC disclosure
β”œβ”€β”€ CONTRIBUTING.md      # Development invariants
└── LICENSE              # MIT

πŸ›‘οΈ Honest Capability Boundaries

Claim Status
Stateful clients across tool calls Yes β€” persistent engine + $RPCMCP
Stateful execution inside a lab VM Yes β€” one persistent guest runspace (rpc_vm_exec); the vmrun fallback is stateless
Context-handle chaining (producer β†’ consumer) Yes β€” store_as / __var__ raw-object passing
Auto-confirm type confusion No β€” NDR cannot prove distinct handle types; verify via RE (see XactSrv case)
Full rogue-RPC hosting No β€” NtObjectManager 2.0.1 ships no server builder; rpc_alpc_squat covers race-capture only
ETW tracing / ALPC SDDL Requires admin (logman / SeDebugPrivilege)
Symbol-resolved procedure names Environment-dependent (symsrv chain); heuristic fallback names otherwise
Runs anywhere but Windows PS 5.1 Not yet (pwsh 7 untested)

πŸ“– Documentation

  • ARCHITECTURE.md β€” engine protocol, state model, design decisions
  • CHANGELOG.md β€” R1–R13 decision history incl. two PS 5.1 marshaling bugs
  • SECURITY.md β€” authorized use, VM isolation, MSRC disclosure
  • CONTRIBUTING.md β€” development invariants and test requirements

⚠️ Disclaimer

For lawful security research, education, and authorized testing only. rpc_call invokes real RPC methods and can crash services β€” run it against an isolated VM, never a production or daily-driver host. Vulnerabilities found through this tool must follow responsible disclosure (MSRC).

πŸ“„ License

MIT β€” Copyright (c) 2026 ntobjmanager-mcp Contributors

Search skills and MCP servers

Fuzzy search across 23,137 skills and servers