diff --git a/.github/mypy/mypy.ini b/.github/mypy/mypy.ini
index e5d862a7..0aded4dc 100644
--- a/.github/mypy/mypy.ini
+++ b/.github/mypy/mypy.ini
@@ -57,6 +57,9 @@ ignore_missing_imports = True
 [mypy-ida_funcs.*]
 ignore_missing_imports = True
 
+[mypy-ida_loader.*]
+ignore_missing_imports = True
+
 [mypy-PyQt5.*]
 ignore_missing_imports = True
 
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 69f05601..65c783eb 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -8,9 +8,14 @@
 - explorer: enforce max column width Features and Editor panes #691 @mike-hunhoff
 - explorer: add option to limit features to currently selected disassembly address #692 @mike-hunhoff
 - all: add support for ELF files #700 @Adir-Shemesh @TcM1911
+- rule format: add feature `format: ` for file format, like `format: pe` #723 @williballenthin
+- rule format: add feature `arch: ` for architecture, like `arch: amd64` #723 @williballenthin
+- rule format: add feature `os: ` for operating system, like `os: windows` #723 @williballenthin
 
 ### Breaking Changes
 
+- legacy term `arch` (i.e., "x32") is now called `bitness` @williballenthin
+
 ### New Rules (24)
 
 - collection/webcam/capture-webcam-image johnk3r
diff --git a/capa/features/common.py b/capa/features/common.py
index 221f15c1..8f16d6e5 100644
--- a/capa/features/common.py
+++ b/capa/features/common.py
@@ -14,6 +14,7 @@ from typing import Set, Dict, Union
 
 import capa.engine
 import capa.features
+import capa.features.extractors.elf
 
 logger = logging.getLogger(__name__)
 MAX_BYTES_FEATURE_SIZE = 0x100
@@ -21,12 +22,6 @@ MAX_BYTES_FEATURE_SIZE = 0x100
 # thunks may be chained so we specify a delta to control the depth to which these chains are explored
 THUNK_CHAIN_DEPTH_DELTA = 5
 
-# identifiers for supported architectures names that tweak a feature
-# for example, offset/x32
-ARCH_X32 = "x32"
-ARCH_X64 = "x64"
-VALID_ARCH = (ARCH_X32, ARCH_X64)
-
 
 def bytes_to_str(b: bytes) -> str:
     return str(codecs.encode(b, "hex").decode("utf-8"))
@@ -52,33 +47,33 @@ def escape_string(s: str) -> str:
 
 
 class Feature:
-    def __init__(self, value: Union[str, int, bytes], arch=None, description=None):
+    def __init__(self, value: Union[str, int, bytes], bitness=None, description=None):
         """
         Args:
           value (any): the value of the feature, such as the number or string.
-          arch (str): one of the VALID_ARCH values, or None.
-            When None, then the feature applies to any architecture.
-            Modifies the feature name from `feature` to `feature/arch`, like `offset/x32`.
+          bitness (str): one of the VALID_BITNESS values, or None.
+            When None, then the feature applies to any bitness.
+            Modifies the feature name from `feature` to `feature/bitness`, like `offset/x32`.
           description (str): a human-readable description that explains the feature value.
         """
         super(Feature, self).__init__()
 
-        if arch is not None:
-            if arch not in VALID_ARCH:
-                raise ValueError("arch '%s' must be one of %s" % (arch, VALID_ARCH))
-            self.name = self.__class__.__name__.lower() + "/" + arch
+        if bitness is not None:
+            if bitness not in VALID_BITNESS:
+                raise ValueError("bitness '%s' must be one of %s" % (bitness, VALID_BITNESS))
+            self.name = self.__class__.__name__.lower() + "/" + bitness
         else:
             self.name = self.__class__.__name__.lower()
 
         self.value = value
-        self.arch = arch
+        self.bitness = bitness
         self.description = description
 
     def __hash__(self):
-        return hash((self.name, self.value, self.arch))
+        return hash((self.name, self.value, self.bitness))
 
     def __eq__(self, other):
-        return self.name == other.name and self.value == other.value and self.arch == other.arch
+        return self.name == other.name and self.value == other.value and self.bitness == other.bitness
 
     def get_value_str(self) -> str:
         """
@@ -105,8 +100,8 @@ class Feature:
         return capa.engine.Result(self in ctx, self, [], locations=ctx.get(self, []))
 
     def freeze_serialize(self):
-        if self.arch is not None:
-            return (self.__class__.__name__, [self.value, {"arch": self.arch}])
+        if self.bitness is not None:
+            return (self.__class__.__name__, [self.value, {"bitness": self.bitness}])
         else:
             return (self.__class__.__name__, [self.value])
 
@@ -131,6 +126,7 @@ class MatchedRule(Feature):
 
 class Characteristic(Feature):
     def __init__(self, value: str, description=None):
+
         super(Characteristic, self).__init__(value, description=description)
 
 
@@ -260,3 +256,57 @@ class Bytes(Feature):
     @classmethod
     def freeze_deserialize(cls, args):
         return cls(*[codecs.decode(x, "hex") for x in args])
+
+
+# identifiers for supported bitness names that tweak a feature
+# for example, offset/x32
+BITNESS_X32 = "x32"
+BITNESS_X64 = "x64"
+VALID_BITNESS = (BITNESS_X32, BITNESS_X64)
+
+
+# other candidates here: https://docs.microsoft.com/en-us/windows/win32/debug/pe-format#machine-types
+ARCH_I386 = "i386"
+ARCH_AMD64 = "amd64"
+VALID_ARCH = (ARCH_I386, ARCH_AMD64)
+
+
+class Arch(Feature):
+    def __init__(self, value: str, description=None):
+        assert value in VALID_ARCH
+        super(Arch, self).__init__(value, description=description)
+        self.name = "arch"
+
+
+OS_WINDOWS = "windows"
+OS_LINUX = "linux"
+OS_MACOS = "macos"
+VALID_OS = {os.value for os in capa.features.extractors.elf.OS}
+VALID_OS.update({OS_WINDOWS, OS_LINUX, OS_MACOS})
+
+
+class OS(Feature):
+    def __init__(self, value: str, description=None):
+        assert value in (VALID_OS)
+        super(OS, self).__init__(value, description=description)
+        self.name = "os"
+
+
+FORMAT_PE = "pe"
+FORMAT_ELF = "elf"
+VALID_FORMAT = (FORMAT_PE, FORMAT_ELF)
+
+
+class Format(Feature):
+    def __init__(self, value: str, description=None):
+        assert value in (VALID_FORMAT)
+        super(Format, self).__init__(value, description=description)
+        self.name = "format"
+
+
+def is_global_feature(feature):
+    """
+    is this a feature that is extracted at every scope?
+    today, these are OS and arch features.
+    """
+    return isinstance(feature, (OS, Arch))
diff --git a/capa/features/extractors/common.py b/capa/features/extractors/common.py
new file mode 100644
index 00000000..450fb636
--- /dev/null
+++ b/capa/features/extractors/common.py
@@ -0,0 +1,78 @@
+import io
+import logging
+import binascii
+import contextlib
+
+import pefile
+
+import capa.features.extractors.elf
+import capa.features.extractors.pefile
+from capa.features.common import OS, FORMAT_PE, FORMAT_ELF, OS_WINDOWS, Arch, Format
+
+logger = logging.getLogger(__name__)
+
+
+def extract_format(buf):
+    if buf.startswith(b"MZ"):
+        yield Format(FORMAT_PE), 0x0
+    elif buf.startswith(b"\x7fELF"):
+        yield Format(FORMAT_ELF), 0x0
+    else:
+        # we likely end up here:
+        #  1. handling a file format (e.g. macho)
+        #
+        # for (1), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s", binascii.hexlify(buf[:4]).decode("ascii"))
+        return
+
+
+def extract_arch(buf):
+    if buf.startswith(b"MZ"):
+        yield from capa.features.extractors.pefile.extract_file_arch(pe=pefile.PE(data=buf))
+
+    elif buf.startswith(b"\x7fELF"):
+        with contextlib.closing(io.BytesIO(buf)) as f:
+            arch = capa.features.extractors.elf.detect_elf_arch(f)
+
+        if arch == "unknown":
+            logger.debug("unsupported arch: %s", arch)
+            return
+
+        yield Arch(arch), 0x0
+
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  2. handling a new file format (e.g. macho)
+        #
+        # for (1) we can't do much - its shellcode and all bets are off.
+        # we could maybe accept a futher CLI argument to specify the arch,
+        # but i think this would be rarely used.
+        # rules that rely on arch conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess Arch", binascii.hexlify(buf[:4]).decode("ascii"))
+        return
+
+
+def extract_os(buf):
+    if buf.startswith(b"MZ"):
+        yield OS(OS_WINDOWS), 0x0
+    elif buf.startswith(b"\x7fELF"):
+        with contextlib.closing(io.BytesIO(buf)) as f:
+            os = capa.features.extractors.elf.detect_elf_os(f)
+
+        yield OS(os), 0x0
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  2. handling a new file format (e.g. macho)
+        #
+        # for (1) we can't do much - its shellcode and all bets are off.
+        # we could maybe accept a futher CLI argument to specify the OS,
+        # but i think this would be rarely used.
+        # rules that rely on OS conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess OS", binascii.hexlify(buf[:4]).decode("ascii"))
+        return
diff --git a/capa/features/extractors/elf.py b/capa/features/extractors/elf.py
new file mode 100644
index 00000000..ce3fe638
--- /dev/null
+++ b/capa/features/extractors/elf.py
@@ -0,0 +1,269 @@
+import struct
+import logging
+from enum import Enum
+from typing import BinaryIO
+
+logger = logging.getLogger(__name__)
+
+
+def align(v, alignment):
+    remainder = v % alignment
+    if remainder == 0:
+        return v
+    else:
+        return v + (alignment - remainder)
+
+
+class CorruptElfFile(ValueError):
+    pass
+
+
+class OS(str, Enum):
+    HPUX = "hpux"
+    NETBSD = "netbsd"
+    LINUX = "linux"
+    HURD = "hurd"
+    _86OPEN = "86open"
+    SOLARIS = "solaris"
+    AIX = "aix"
+    IRIX = "irix"
+    FREEBSD = "freebsd"
+    TRU64 = "tru64"
+    MODESTO = "modesto"
+    OPENBSD = "openbsd"
+    OPENVMS = "openvms"
+    NSK = "nsk"
+    AROS = "aros"
+    FENIXOS = "fenixos"
+    CLOUD = "cloud"
+    SYLLABLE = "syllable"
+    NACL = "nacl"
+
+
+def detect_elf_os(f: BinaryIO) -> str:
+    f.seek(0x0)
+    file_header = f.read(0x40)
+
+    # we'll set this to the detected OS
+    # prefer the first heuristics,
+    # but rather than short circuiting,
+    # we'll still parse out the remainder, for debugging.
+    ret = None
+
+    if not file_header.startswith(b"\x7fELF"):
+        raise CorruptElfFile("missing magic header")
+
+    ei_class, ei_data = struct.unpack_from("BB", file_header, 4)
+    logger.debug("ei_class: 0x%02x ei_data: 0x%02x", ei_class, ei_data)
+    if ei_class == 1:
+        bitness = 32
+    elif ei_class == 2:
+        bitness = 64
+    else:
+        raise CorruptElfFile("invalid ei_class: 0x%02x" % ei_class)
+
+    if ei_data == 1:
+        endian = "<"
+    elif ei_data == 2:
+        endian = ">"
+    else:
+        raise CorruptElfFile("not an ELF file: invalid ei_data: 0x%02x" % ei_data)
+
+    if bitness == 32:
+        (e_phoff,) = struct.unpack_from(endian + "I", file_header, 0x1C)
+        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x2A)
+    elif bitness == 64:
+        (e_phoff,) = struct.unpack_from(endian + "Q", file_header, 0x20)
+        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x36)
+    else:
+        raise NotImplemented
+
+    logger.debug("e_phoff: 0x%02x e_phentsize: 0x%02x e_phnum: %d", e_phoff, e_phentsize, e_phnum)
+
+    (ei_osabi,) = struct.unpack_from(endian + "B", file_header, 7)
+    OSABI = {
+        # via pyelftools: https://github.com/eliben/pyelftools/blob/0664de05ed2db3d39041e2d51d19622a8ef4fb0f/elftools/elf/enums.py#L35-L58
+        # some candidates are commented out because the are not useful values,
+        # at least when guessing OSes
+        # 0: "SYSV",  # too often used when OS is not SYSV
+        1: OS.HPUX,
+        2: OS.NETBSD,
+        3: OS.LINUX,
+        4: OS.HURD,
+        5: OS._86OPEN,
+        6: OS.SOLARIS,
+        7: OS.AIX,
+        8: OS.IRIX,
+        9: OS.FREEBSD,
+        10: OS.TRU64,
+        11: OS.MODESTO,
+        12: OS.OPENBSD,
+        13: OS.OPENVMS,
+        14: OS.NSK,
+        15: OS.AROS,
+        16: OS.FENIXOS,
+        17: OS.CLOUD,
+        # 53: "SORTFIX",      # i can't find any reference to this OS, i dont think it exists
+        # 64: "ARM_AEABI",    # not an OS
+        # 97: "ARM",          # not an OS
+        # 255: "STANDALONE",  # not an OS
+    }
+    logger.debug("ei_osabi: 0x%02x (%s)", ei_osabi, OSABI.get(ei_osabi, "unknown"))
+
+    # os_osabi == 0 is commonly set even when the OS is not SYSV.
+    # other values are unused or unknown.
+    if ei_osabi in OSABI and ei_osabi != 0x0:
+        # subsequent strategies may overwrite this value
+        ret = OSABI[ei_osabi]
+
+    f.seek(e_phoff)
+    program_header_size = e_phnum * e_phentsize
+    program_headers = f.read(program_header_size)
+    if len(program_headers) != program_header_size:
+        logger.warning("failed to read program headers")
+        e_phnum = 0
+
+    # search for PT_NOTE sections that specify an OS
+    # for example, on Linux there is a GNU section with minimum kernel version
+    for i in range(e_phnum):
+        offset = i * e_phentsize
+        phent = program_headers[offset : offset + e_phentsize]
+
+        PT_NOTE = 0x4
+
+        (p_type,) = struct.unpack_from(endian + "I", phent, 0x0)
+        logger.debug("p_type: 0x%04x", p_type)
+        if p_type != PT_NOTE:
+            continue
+
+        if bitness == 32:
+            p_offset, _, _, p_filesz = struct.unpack_from(endian + "IIII", phent, 0x4)
+        elif bitness == 64:
+            p_offset, _, _, p_filesz = struct.unpack_from(endian + "QQQQ", phent, 0x8)
+        else:
+            raise NotImplemented
+
+        logger.debug("p_offset: 0x%02x p_filesz: 0x%04x", p_offset, p_filesz)
+
+        f.seek(p_offset)
+        note = f.read(p_filesz)
+        if len(note) != p_filesz:
+            logger.warning("failed to read note content")
+            continue
+
+        namesz, descsz, type_ = struct.unpack_from(endian + "III", note, 0x0)
+        name_offset = 0xC
+        desc_offset = name_offset + align(namesz, 0x4)
+
+        logger.debug("namesz: 0x%02x descsz: 0x%02x type: 0x%04x", namesz, descsz, type_)
+
+        name = note[name_offset : name_offset + namesz].partition(b"\x00")[0].decode("ascii")
+        logger.debug("name: %s", name)
+
+        if type_ != 1:
+            continue
+
+        if name == "GNU":
+            if descsz < 16:
+                continue
+
+            desc = note[desc_offset : desc_offset + descsz]
+            abi_tag, kmajor, kminor, kpatch = struct.unpack_from(endian + "IIII", desc, 0x0)
+            # via readelf: https://github.com/bminor/binutils-gdb/blob/c0e94211e1ac05049a4ce7c192c9d14d1764eb3e/binutils/readelf.c#L19635-L19658
+            # and here: https://github.com/bminor/binutils-gdb/blob/34c54daa337da9fadf87d2706d6a590ae1f88f4d/include/elf/common.h#L933-L939
+            GNU_ABI_TAG = {
+                0: OS.LINUX,
+                1: OS.HURD,
+                2: OS.SOLARIS,
+                3: OS.FREEBSD,
+                4: OS.NETBSD,
+                5: OS.SYLLABLE,
+                6: OS.NACL,
+            }
+            logger.debug("GNU_ABI_TAG: 0x%02x", abi_tag)
+
+            if abi_tag in GNU_ABI_TAG:
+                # update only if not set
+                # so we can get the debugging output of subsequent strategies
+                ret = GNU_ABI_TAG[abi_tag] if not ret else ret
+                logger.debug("abi tag: %s earliest compatible kernel: %d.%d.%d", ret, kmajor, kminor, kpatch)
+        elif name == "OpenBSD":
+            logger.debug("note owner: %s", "OPENBSD")
+            ret = OS.OPENBSD if not ret else ret
+        elif name == "NetBSD":
+            logger.debug("note owner: %s", "NETBSD")
+            ret = OS.NETBSD if not ret else ret
+        elif name == "FreeBSD":
+            logger.debug("note owner: %s", "FREEBSD")
+            ret = OS.FREEBSD if not ret else ret
+
+    # search for recognizable dynamic linkers (interpreters)
+    # for example, on linux, we see file paths like: /lib64/ld-linux-x86-64.so.2
+    for i in range(e_phnum):
+        offset = i * e_phentsize
+        phent = program_headers[offset : offset + e_phentsize]
+
+        PT_INTERP = 0x3
+
+        (p_type,) = struct.unpack_from(endian + "I", phent, 0x0)
+        if p_type != PT_INTERP:
+            continue
+
+        if bitness == 32:
+            p_offset, _, _, p_filesz = struct.unpack_from(endian + "IIII", phent, 0x4)
+        elif bitness == 64:
+            p_offset, _, _, p_filesz = struct.unpack_from(endian + "QQQQ", phent, 0x8)
+        else:
+            raise NotImplemented
+
+        f.seek(p_offset)
+        interp = f.read(p_filesz)
+        if len(interp) != p_filesz:
+            logger.warning("failed to read interp content")
+            continue
+
+        linker = interp.partition(b"\x00")[0].decode("ascii")
+        logger.debug("linker: %s", linker)
+        if "ld-linux" in linker:
+            # update only if not set
+            # so we can get the debugging output of subsequent strategies
+            ret = OS.LINUX if ret is None else ret
+
+    return ret.value if ret is not None else "unknown"
+
+
+class Arch(str, Enum):
+    I386 = "i386"
+    AMD64 = "amd64"
+
+
+def detect_elf_arch(f: BinaryIO) -> str:
+    f.seek(0x0)
+    file_header = f.read(0x40)
+
+    if not file_header.startswith(b"\x7fELF"):
+        raise CorruptElfFile("missing magic header")
+
+    (ei_data,) = struct.unpack_from("B", file_header, 5)
+    logger.debug("ei_data: 0x%02x", ei_data)
+
+    if ei_data == 1:
+        endian = "<"
+    elif ei_data == 2:
+        endian = ">"
+    else:
+        raise CorruptElfFile("not an ELF file: invalid ei_data: 0x%02x" % ei_data)
+
+    (ei_machine,) = struct.unpack_from(endian + "H", file_header, 0x12)
+    logger.debug("ei_machine: 0x%02x", ei_machine)
+
+    EM_386 = 0x3
+    EM_X86_64 = 0x3E
+    if ei_machine == EM_386:
+        return Arch.I386
+    elif ei_machine == EM_X86_64:
+        return Arch.AMD64
+    else:
+        # not really unknown, but unsupport at the moment:
+        # https://github.com/eliben/pyelftools/blob/ab444d982d1849191e910299a985989857466620/elftools/elf/enums.py#L73
+        return "unknown"
diff --git a/capa/features/extractors/ida/extractor.py b/capa/features/extractors/ida/extractor.py
index dc260adc..7d00c7ef 100644
--- a/capa/features/extractors/ida/extractor.py
+++ b/capa/features/extractors/ida/extractor.py
@@ -5,11 +5,13 @@
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-
 import idaapi
 
+import capa.ida.helpers
+import capa.features.extractors.elf
 import capa.features.extractors.ida.file
 import capa.features.extractors.ida.insn
+import capa.features.extractors.ida.global_
 import capa.features.extractors.ida.function
 import capa.features.extractors.ida.basicblock
 from capa.features.extractors.base_extractor import FeatureExtractor
@@ -57,6 +59,9 @@ class InstructionHandle:
 class IdaFeatureExtractor(FeatureExtractor):
     def __init__(self):
         super(IdaFeatureExtractor, self).__init__()
+        self.global_features = []
+        self.global_features.extend(capa.features.extractors.ida.global_.extract_os())
+        self.global_features.extend(capa.features.extractors.ida.global_.extract_arch())
 
     def get_base_address(self):
         return idaapi.get_imagebase()
@@ -64,6 +69,7 @@ class IdaFeatureExtractor(FeatureExtractor):
     def extract_file_features(self):
         for (feature, ea) in capa.features.extractors.ida.file.extract_features():
             yield feature, ea
+        yield from self.global_features
 
     def get_functions(self):
         import capa.features.extractors.ida.helpers as ida_helpers
@@ -86,6 +92,7 @@ class IdaFeatureExtractor(FeatureExtractor):
     def extract_function_features(self, f):
         for (feature, ea) in capa.features.extractors.ida.function.extract_features(f):
             yield feature, ea
+        yield from self.global_features
 
     def get_basic_blocks(self, f):
         import capa.features.extractors.ida.helpers as ida_helpers
@@ -96,6 +103,7 @@ class IdaFeatureExtractor(FeatureExtractor):
     def extract_basic_block_features(self, f, bb):
         for (feature, ea) in capa.features.extractors.ida.basicblock.extract_features(f, bb):
             yield feature, ea
+        yield from self.global_features
 
     def get_instructions(self, f, bb):
         import capa.features.extractors.ida.helpers as ida_helpers
@@ -106,3 +114,4 @@ class IdaFeatureExtractor(FeatureExtractor):
     def extract_insn_features(self, f, bb, insn):
         for (feature, ea) in capa.features.extractors.ida.insn.extract_features(f, bb, insn):
             yield feature, ea
+        yield from self.global_features
diff --git a/capa/features/extractors/ida/file.py b/capa/features/extractors/ida/file.py
index d3f6a101..3fa72351 100644
--- a/capa/features/extractors/ida/file.py
+++ b/capa/features/extractors/ida/file.py
@@ -11,12 +11,13 @@ import struct
 import idc
 import idaapi
 import idautils
+import ida_loader
 
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 import capa.features.extractors.ida.helpers
 from capa.features.file import Export, Import, Section, FunctionName
-from capa.features.common import String, Characteristic
+from capa.features.common import OS, FORMAT_PE, FORMAT_ELF, OS_WINDOWS, Format, String, Characteristic
 
 
 def check_segment_for_pe(seg):
@@ -153,6 +154,19 @@ def extract_file_function_names():
             yield FunctionName(name), ea
 
 
+def extract_file_format():
+    format_name = ida_loader.get_file_type_name()
+
+    if "PE" in format_name:
+        yield Format(FORMAT_PE), 0x0
+    elif "ELF64" in format_name:
+        yield Format(FORMAT_ELF), 0x0
+    elif "ELF32" in format_name:
+        yield Format(FORMAT_ELF), 0x0
+    else:
+        raise NotImplementedError("file format: %s", format_name)
+
+
 def extract_features():
     """extract file features"""
     for file_handler in FILE_HANDLERS:
@@ -167,6 +181,7 @@ FILE_HANDLERS = (
     extract_file_section_names,
     extract_file_embedded_pe,
     extract_file_function_names,
+    extract_file_format,
 )
 
 
diff --git a/capa/features/extractors/ida/global_.py b/capa/features/extractors/ida/global_.py
new file mode 100644
index 00000000..b3181293
--- /dev/null
+++ b/capa/features/extractors/ida/global_.py
@@ -0,0 +1,56 @@
+import logging
+import contextlib
+
+import idaapi
+import ida_loader
+
+import capa.ida.helpers
+import capa.features.extractors.elf
+from capa.features.common import OS, ARCH_I386, ARCH_AMD64, OS_WINDOWS, Arch
+
+logger = logging.getLogger(__name__)
+
+
+def extract_os():
+    format_name = ida_loader.get_file_type_name()
+
+    if "PE" in format_name:
+        yield OS(OS_WINDOWS), 0x0
+
+    elif "ELF" in format_name:
+        with contextlib.closing(capa.ida.helpers.IDAIO()) as f:
+            os = capa.features.extractors.elf.detect_elf_os(f)
+
+        yield OS(os), 0x0
+
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  2. handling a new file format (e.g. macho)
+        #
+        # for (1) we can't do much - its shellcode and all bets are off.
+        # we could maybe accept a futher CLI argument to specify the OS,
+        # but i think this would be rarely used.
+        # rules that rely on OS conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess OS", format_name)
+        return
+
+
+def extract_arch():
+    info = idaapi.get_inf_structure()
+    if info.procName == "metapc" and info.is_64bit():
+        yield Arch(ARCH_AMD64), 0x0
+    elif info.procName == "metapc" and info.is_32bit():
+        yield Arch(ARCH_I386), 0x0
+    elif info.procName == "metapc":
+        logger.debug("unsupported architecture: non-32-bit nor non-64-bit intel")
+        return
+    else:
+        # we likely end up here:
+        #  1. handling a new architecture (e.g. aarch64)
+        #
+        # for (1), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported architecture: %s", info.procName)
+        return
diff --git a/capa/features/extractors/ida/insn.py b/capa/features/extractors/ida/insn.py
index 82fb4715..8bfcd7fb 100644
--- a/capa/features/extractors/ida/insn.py
+++ b/capa/features/extractors/ida/insn.py
@@ -14,8 +14,8 @@ import capa.features.extractors.helpers
 import capa.features.extractors.ida.helpers
 from capa.features.insn import API, Number, Offset, Mnemonic
 from capa.features.common import (
-    ARCH_X32,
-    ARCH_X64,
+    BITNESS_X32,
+    BITNESS_X64,
     MAX_BYTES_FEATURE_SIZE,
     THUNK_CHAIN_DEPTH_DELTA,
     Bytes,
@@ -28,22 +28,22 @@ from capa.features.common import (
 SECURITY_COOKIE_BYTES_DELTA = 0x40
 
 
-def get_arch(ctx):
+def get_bitness(ctx):
     """
-    fetch the ARCH_* constant for the currently open workspace.
+    fetch the BITNESS_* constant for the currently open workspace.
 
     via Tamir Bahar/@tmr232
     https://reverseengineering.stackexchange.com/a/11398/17194
     """
-    if "arch" not in ctx:
+    if "bitness" not in ctx:
         info = idaapi.get_inf_structure()
         if info.is_64bit():
-            ctx["arch"] = ARCH_X64
+            ctx["bitness"] = BITNESS_X64
         elif info.is_32bit():
-            ctx["arch"] = ARCH_X32
+            ctx["bitness"] = BITNESS_X32
         else:
-            raise ValueError("unexpected architecture")
-    return ctx["arch"]
+            raise ValueError("unexpected bitness")
+    return ctx["bitness"]
 
 
 def get_imports(ctx):
@@ -149,7 +149,7 @@ def extract_insn_number_features(f, bb, insn):
             const = op.addr
 
         yield Number(const), insn.ea
-        yield Number(const, arch=get_arch(f.ctx)), insn.ea
+        yield Number(const, bitness=get_bitness(f.ctx)), insn.ea
 
 
 def extract_insn_bytes_features(f, bb, insn):
@@ -218,7 +218,7 @@ def extract_insn_offset_features(f, bb, insn):
         op_off = capa.features.extractors.helpers.twos_complement(op_off, 32)
 
         yield Offset(op_off), insn.ea
-        yield Offset(op_off, arch=get_arch(f.ctx)), insn.ea
+        yield Offset(op_off, bitness=get_bitness(f.ctx)), insn.ea
 
 
 def contains_stack_cookie_keywords(s):
diff --git a/capa/features/extractors/pefile.py b/capa/features/extractors/pefile.py
index e27d6c20..3e5d97fd 100644
--- a/capa/features/extractors/pefile.py
+++ b/capa/features/extractors/pefile.py
@@ -9,25 +9,23 @@ import logging
 
 import pefile
 
+import capa.features.common
 import capa.features.extractors
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 from capa.features.file import Export, Import, Section
-from capa.features.common import String, Characteristic
+from capa.features.common import OS, ARCH_I386, FORMAT_PE, ARCH_AMD64, OS_WINDOWS, Arch, Format, String, Characteristic
 from capa.features.extractors.base_extractor import FeatureExtractor
 
 logger = logging.getLogger(__name__)
 
 
-def extract_file_embedded_pe(pe, file_path):
-    with open(file_path, "rb") as f:
-        fbytes = f.read()
-
-    for offset, i in capa.features.extractors.helpers.carve_pe(fbytes, 1):
+def extract_file_embedded_pe(buf, **kwargs):
+    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
         yield Characteristic("embedded pe"), offset
 
 
-def extract_file_export_names(pe, file_path):
+def extract_file_export_names(pe, **kwargs):
     base_address = pe.OPTIONAL_HEADER.ImageBase
 
     if hasattr(pe, "DIRECTORY_ENTRY_EXPORT"):
@@ -42,7 +40,7 @@ def extract_file_export_names(pe, file_path):
             yield Export(name), va
 
 
-def extract_file_import_names(pe, file_path):
+def extract_file_import_names(pe, **kwargs):
     """
     extract imported function names
     1. imports by ordinal:
@@ -74,7 +72,7 @@ def extract_file_import_names(pe, file_path):
                     yield Import(name), imp.address
 
 
-def extract_file_section_names(pe, file_path):
+def extract_file_section_names(pe, **kwargs):
     base_address = pe.OPTIONAL_HEADER.ImageBase
 
     for section in pe.sections:
@@ -86,21 +84,18 @@ def extract_file_section_names(pe, file_path):
         yield Section(name), base_address + section.VirtualAddress
 
 
-def extract_file_strings(pe, file_path):
+def extract_file_strings(buf, **kwargs):
     """
     extract ASCII and UTF-16 LE strings from file
     """
-    with open(file_path, "rb") as f:
-        b = f.read()
-
-    for s in capa.features.extractors.strings.extract_ascii_strings(b):
+    for s in capa.features.extractors.strings.extract_ascii_strings(buf):
         yield String(s.s), s.offset
 
-    for s in capa.features.extractors.strings.extract_unicode_strings(b):
+    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
         yield String(s.s), s.offset
 
 
-def extract_file_function_names(pe, file_path):
+def extract_file_function_names(**kwargs):
     """
     extract the names of statically-linked library functions.
     """
@@ -110,20 +105,39 @@ def extract_file_function_names(pe, file_path):
     return
 
 
-def extract_file_features(pe, file_path):
+def extract_file_os(**kwargs):
+    # assuming PE -> Windows
+    # though i suppose they're also used by UEFI
+    yield OS(OS_WINDOWS), 0x0
+
+
+def extract_file_format(**kwargs):
+    yield Format(FORMAT_PE), 0x0
+
+
+def extract_file_arch(pe, **kwargs):
+    if pe.FILE_HEADER.Machine == pefile.MACHINE_TYPE["IMAGE_FILE_MACHINE_I386"]:
+        yield Arch(ARCH_I386), 0x0
+    elif pe.FILE_HEADER.Machine == pefile.MACHINE_TYPE["IMAGE_FILE_MACHINE_AMD64"]:
+        yield Arch(ARCH_AMD64), 0x0
+    else:
+        logger.warning("unsupported architecture: %s", pefile.MACHINE_TYPE[pe.FILE_HEADER.Machine])
+
+
+def extract_file_features(pe, buf):
     """
     extract file features from given workspace
 
     args:
       pe (pefile.PE): the parsed PE
-      file_path: path to the input file
+      buf: the raw sample bytes
 
     yields:
       Tuple[Feature, VA]: a feature and its location.
     """
 
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(pe, file_path):
+        for feature, va in file_handler(pe=pe, buf=buf):
             yield feature, va
 
 
@@ -134,6 +148,9 @@ FILE_HANDLERS = (
     extract_file_section_names,
     extract_file_strings,
     extract_file_function_names,
+    extract_file_os,
+    extract_file_format,
+    extract_file_arch,
 )
 
 
@@ -147,7 +164,10 @@ class PefileFeatureExtractor(FeatureExtractor):
         return self.pe.OPTIONAL_HEADER.ImageBase
 
     def extract_file_features(self):
-        for feature, va in extract_file_features(self.pe, self.path):
+        with open(self.path, "rb") as f:
+            buf = f.read()
+
+        for feature, va in extract_file_features(self.pe, buf):
             yield feature, va
 
     def get_functions(self):
diff --git a/capa/features/extractors/smda/extractor.py b/capa/features/extractors/smda/extractor.py
index b4355d8f..1a13653b 100644
--- a/capa/features/extractors/smda/extractor.py
+++ b/capa/features/extractors/smda/extractor.py
@@ -1,7 +1,9 @@
 from smda.common.SmdaReport import SmdaReport
 
+import capa.features.extractors.common
 import capa.features.extractors.smda.file
 import capa.features.extractors.smda.insn
+import capa.features.extractors.smda.global_
 import capa.features.extractors.smda.function
 import capa.features.extractors.smda.basicblock
 from capa.features.extractors.base_extractor import FeatureExtractor
@@ -12,13 +14,21 @@ class SmdaFeatureExtractor(FeatureExtractor):
         super(SmdaFeatureExtractor, self).__init__()
         self.smda_report = smda_report
         self.path = path
+        with open(self.path, "rb") as f:
+            self.buf = f.read()
+
+        # pre-compute these because we'll yield them at *every* scope.
+        self.global_features = []
+        self.global_features.extend(capa.features.extractors.common.extract_os(self.buf))
+        self.global_features.extend(capa.features.extractors.smda.global_.extract_arch(self.smda_report))
 
     def get_base_address(self):
         return self.smda_report.base_addr
 
     def extract_file_features(self):
-        for feature, va in capa.features.extractors.smda.file.extract_features(self.smda_report, self.path):
+        for feature, va in capa.features.extractors.smda.file.extract_features(self.smda_report, self.buf):
             yield feature, va
+        yield from self.global_features
 
     def get_functions(self):
         for function in self.smda_report.getFunctions():
@@ -27,6 +37,7 @@ class SmdaFeatureExtractor(FeatureExtractor):
     def extract_function_features(self, f):
         for feature, va in capa.features.extractors.smda.function.extract_features(f):
             yield feature, va
+        yield from self.global_features
 
     def get_basic_blocks(self, f):
         for bb in f.getBlocks():
@@ -35,6 +46,7 @@ class SmdaFeatureExtractor(FeatureExtractor):
     def extract_basic_block_features(self, f, bb):
         for feature, va in capa.features.extractors.smda.basicblock.extract_features(f, bb):
             yield feature, va
+        yield from self.global_features
 
     def get_instructions(self, f, bb):
         for smda_ins in bb.getInstructions():
@@ -43,3 +55,4 @@ class SmdaFeatureExtractor(FeatureExtractor):
     def extract_insn_features(self, f, bb, insn):
         for feature, va in capa.features.extractors.smda.insn.extract_features(f, bb, insn):
             yield feature, va
+        yield from self.global_features
diff --git a/capa/features/extractors/smda/file.py b/capa/features/extractors/smda/file.py
index 5250e26b..fe901bea 100644
--- a/capa/features/extractors/smda/file.py
+++ b/capa/features/extractors/smda/file.py
@@ -1,30 +1,29 @@
 # if we have SMDA we definitely have lief
 import lief
 
+import capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 from capa.features.file import Export, Import, Section
 from capa.features.common import String, Characteristic
 
 
-def extract_file_embedded_pe(smda_report, file_path):
-    with open(file_path, "rb") as f:
-        fbytes = f.read()
-
-    for offset, i in capa.features.extractors.helpers.carve_pe(fbytes, 1):
+def extract_file_embedded_pe(buf, **kwargs):
+    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
         yield Characteristic("embedded pe"), offset
 
 
-def extract_file_export_names(smda_report, file_path):
-    lief_binary = lief.parse(file_path)
+def extract_file_export_names(buf, **kwargs):
+    lief_binary = lief.parse(buf)
+
     if lief_binary is not None:
         for function in lief_binary.exported_functions:
             yield Export(function.name), function.address
 
 
-def extract_file_import_names(smda_report, file_path):
+def extract_file_import_names(smda_report, buf):
     # extract import table info via LIEF
-    lief_binary = lief.parse(file_path)
+    lief_binary = lief.parse(buf)
     if not isinstance(lief_binary, lief.PE.Binary):
         return
     for imported_library in lief_binary.imports:
@@ -40,8 +39,8 @@ def extract_file_import_names(smda_report, file_path):
                     yield Import(name), va
 
 
-def extract_file_section_names(smda_report, file_path):
-    lief_binary = lief.parse(file_path)
+def extract_file_section_names(buf, **kwargs):
+    lief_binary = lief.parse(buf)
     if not isinstance(lief_binary, lief.PE.Binary):
         return
     if lief_binary and lief_binary.sections:
@@ -50,21 +49,18 @@ def extract_file_section_names(smda_report, file_path):
             yield Section(section.name), base_address + section.virtual_address
 
 
-def extract_file_strings(smda_report, file_path):
+def extract_file_strings(buf, **kwargs):
     """
     extract ASCII and UTF-16 LE strings from file
     """
-    with open(file_path, "rb") as f:
-        b = f.read()
-
-    for s in capa.features.extractors.strings.extract_ascii_strings(b):
+    for s in capa.features.extractors.strings.extract_ascii_strings(buf):
         yield String(s.s), s.offset
 
-    for s in capa.features.extractors.strings.extract_unicode_strings(b):
+    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
         yield String(s.s), s.offset
 
 
-def extract_file_function_names(smda_report, file_path):
+def extract_file_function_names(smda_report, **kwargs):
     """
     extract the names of statically-linked library functions.
     """
@@ -74,20 +70,24 @@ def extract_file_function_names(smda_report, file_path):
     return
 
 
-def extract_features(smda_report, file_path):
+def extract_file_format(buf, **kwargs):
+    yield from capa.features.extractors.common.extract_format(buf)
+
+
+def extract_features(smda_report, buf):
     """
     extract file features from given workspace
 
     args:
       smda_report (smda.common.SmdaReport): a SmdaReport
-      file_path: path to the input file
+      buf: the raw bytes of the sample
 
     yields:
       Tuple[Feature, VA]: a feature and its location.
     """
 
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(smda_report, file_path):
+        for feature, va in file_handler(smda_report=smda_report, buf=buf):
             yield feature, va
 
 
@@ -98,4 +98,5 @@ FILE_HANDLERS = (
     extract_file_section_names,
     extract_file_strings,
     extract_file_function_names,
+    extract_file_format,
 )
diff --git a/capa/features/extractors/smda/global_.py b/capa/features/extractors/smda/global_.py
new file mode 100644
index 00000000..84befc59
--- /dev/null
+++ b/capa/features/extractors/smda/global_.py
@@ -0,0 +1,20 @@
+import logging
+
+from capa.features.common import ARCH_I386, ARCH_AMD64, Arch
+
+logger = logging.getLogger(__name__)
+
+
+def extract_arch(smda_report):
+    if smda_report.architecture == "intel":
+        if smda_report.bitness == 32:
+            yield Arch(ARCH_I386), 0x0
+        elif smda_report.bitness == 64:
+            yield Arch(ARCH_AMD64), 0x0
+    else:
+        # we likely end up here:
+        #  1. handling a new architecture (e.g. aarch64)
+        #
+        # for (1), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported architecture: %s", smda_report.architecture)
+        return
diff --git a/capa/features/extractors/smda/insn.py b/capa/features/extractors/smda/insn.py
index da5ebbfa..d666728c 100644
--- a/capa/features/extractors/smda/insn.py
+++ b/capa/features/extractors/smda/insn.py
@@ -7,8 +7,8 @@ from smda.common.SmdaReport import SmdaReport
 import capa.features.extractors.helpers
 from capa.features.insn import API, Number, Offset, Mnemonic
 from capa.features.common import (
-    ARCH_X32,
-    ARCH_X64,
+    BITNESS_X32,
+    BITNESS_X64,
     MAX_BYTES_FEATURE_SIZE,
     THUNK_CHAIN_DEPTH_DELTA,
     Bytes,
@@ -23,12 +23,12 @@ PATTERN_HEXNUM = re.compile(r"[+\-] (?P<num>0x[a-fA-F0-9]+)")
 PATTERN_SINGLENUM = re.compile(r"[+\-] (?P<num>[0-9])")
 
 
-def get_arch(smda_report):
+def get_bitness(smda_report):
     if smda_report.architecture == "intel":
         if smda_report.bitness == 32:
-            return ARCH_X32
+            return BITNESS_X32
         elif smda_report.bitness == 64:
-            return ARCH_X64
+            return BITNESS_X64
     else:
         raise NotImplementedError
 
@@ -85,7 +85,7 @@ def extract_insn_number_features(f, bb, insn):
     for operand in operands:
         try:
             yield Number(int(operand, 16)), insn.offset
-            yield Number(int(operand, 16), arch=get_arch(f.smda_report)), insn.offset
+            yield Number(int(operand, 16), bitness=get_bitness(f.smda_report)), insn.offset
         except:
             continue
 
@@ -228,7 +228,7 @@ def extract_insn_offset_features(f, bb, insn):
             number = int(number_int.group("num"))
             number = -1 * number if number_int.group().startswith("-") else number
         yield Offset(number), insn.offset
-        yield Offset(number, arch=get_arch(f.smda_report)), insn.offset
+        yield Offset(number, bitness=get_bitness(f.smda_report)), insn.offset
 
 
 def is_security_cookie(f, bb, insn):
diff --git a/capa/features/extractors/viv/extractor.py b/capa/features/extractors/viv/extractor.py
index 15f3973a..b18d4349 100644
--- a/capa/features/extractors/viv/extractor.py
+++ b/capa/features/extractors/viv/extractor.py
@@ -10,8 +10,10 @@ import logging
 import viv_utils
 import viv_utils.flirt
 
+import capa.features.extractors.common
 import capa.features.extractors.viv.file
 import capa.features.extractors.viv.insn
+import capa.features.extractors.viv.global_
 import capa.features.extractors.viv.function
 import capa.features.extractors.viv.basicblock
 from capa.features.extractors.base_extractor import FeatureExtractor
@@ -37,14 +39,22 @@ class VivisectFeatureExtractor(FeatureExtractor):
         super(VivisectFeatureExtractor, self).__init__()
         self.vw = vw
         self.path = path
+        with open(self.path, "rb") as f:
+            self.buf = f.read()
+
+        # pre-compute these because we'll yield them at *every* scope.
+        self.global_features = []
+        self.global_features.extend(capa.features.extractors.common.extract_os(self.buf))
+        self.global_features.extend(capa.features.extractors.viv.global_.extract_arch(self.vw))
 
     def get_base_address(self):
         # assume there is only one file loaded into the vw
         return list(self.vw.filemeta.values())[0]["imagebase"]
 
     def extract_file_features(self):
-        for feature, va in capa.features.extractors.viv.file.extract_features(self.vw, self.path):
+        for feature, va in capa.features.extractors.viv.file.extract_features(self.vw, self.buf):
             yield feature, va
+        yield from self.global_features
 
     def get_functions(self):
         for va in sorted(self.vw.getFunctions()):
@@ -53,6 +63,7 @@ class VivisectFeatureExtractor(FeatureExtractor):
     def extract_function_features(self, f):
         for feature, va in capa.features.extractors.viv.function.extract_features(f):
             yield feature, va
+        yield from self.global_features
 
     def get_basic_blocks(self, f):
         return f.basic_blocks
@@ -60,6 +71,7 @@ class VivisectFeatureExtractor(FeatureExtractor):
     def extract_basic_block_features(self, f, bb):
         for feature, va in capa.features.extractors.viv.basicblock.extract_features(f, bb):
             yield feature, va
+        yield from self.global_features
 
     def get_instructions(self, f, bb):
         for insn in bb.instructions:
@@ -68,6 +80,7 @@ class VivisectFeatureExtractor(FeatureExtractor):
     def extract_insn_features(self, f, bb, insn):
         for feature, va in capa.features.extractors.viv.insn.extract_features(f, bb, insn):
             yield feature, va
+        yield from self.global_features
 
     def is_library_function(self, va):
         return viv_utils.flirt.is_library_function(self.vw, va)
diff --git a/capa/features/extractors/viv/file.py b/capa/features/extractors/viv/file.py
index 00eb57ed..4b9cd13f 100644
--- a/capa/features/extractors/viv/file.py
+++ b/capa/features/extractors/viv/file.py
@@ -11,26 +11,24 @@ import viv_utils
 import viv_utils.flirt
 
 import capa.features.insn
+import capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 from capa.features.file import Export, Import, Section, FunctionName
 from capa.features.common import String, Characteristic
 
 
-def extract_file_embedded_pe(vw, file_path):
-    with open(file_path, "rb") as f:
-        fbytes = f.read()
-
-    for offset, i in pe_carve.carve(fbytes, 1):
+def extract_file_embedded_pe(buf, **kwargs):
+    for offset, _ in pe_carve.carve(buf, 1):
         yield Characteristic("embedded pe"), offset
 
 
-def extract_file_export_names(vw, file_path):
-    for va, etype, name, _ in vw.getExports():
+def extract_file_export_names(vw, **kwargs):
+    for va, _, name, _ in vw.getExports():
         yield Export(name), va
 
 
-def extract_file_import_names(vw, file_path):
+def extract_file_import_names(vw, **kwargs):
     """
     extract imported function names
     1. imports by ordinal:
@@ -64,26 +62,23 @@ def is_viv_ord_impname(impname: str) -> bool:
         return True
 
 
-def extract_file_section_names(vw, file_path):
+def extract_file_section_names(vw, **kwargs):
     for va, _, segname, _ in vw.getSegments():
         yield Section(segname), va
 
 
-def extract_file_strings(vw, file_path):
+def extract_file_strings(buf, **kwargs):
     """
     extract ASCII and UTF-16 LE strings from file
     """
-    with open(file_path, "rb") as f:
-        b = f.read()
-
-    for s in capa.features.extractors.strings.extract_ascii_strings(b):
+    for s in capa.features.extractors.strings.extract_ascii_strings(buf):
         yield String(s.s), s.offset
 
-    for s in capa.features.extractors.strings.extract_unicode_strings(b):
+    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
         yield String(s.s), s.offset
 
 
-def extract_file_function_names(vw, file_path):
+def extract_file_function_names(vw, **kwargs):
     """
     extract the names of statically-linked library functions.
     """
@@ -93,20 +88,24 @@ def extract_file_function_names(vw, file_path):
             yield FunctionName(name), va
 
 
-def extract_features(vw, file_path):
+def extract_file_format(buf, **kwargs):
+    yield from capa.features.extractors.common.extract_format(buf)
+
+
+def extract_features(vw, buf: bytes):
     """
     extract file features from given workspace
 
     args:
       vw (vivisect.VivWorkspace): the vivisect workspace
-      file_path: path to the input file
+      buf: the raw input file bytes
 
     yields:
       Tuple[Feature, VA]: a feature and its location.
     """
 
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(vw, file_path):
+        for feature, va in file_handler(vw=vw, buf=buf):  # type: ignore
             yield feature, va
 
 
@@ -117,4 +116,5 @@ FILE_HANDLERS = (
     extract_file_section_names,
     extract_file_strings,
     extract_file_function_names,
+    extract_file_format,
 )
diff --git a/capa/features/extractors/viv/global_.py b/capa/features/extractors/viv/global_.py
new file mode 100644
index 00000000..8fc08ee2
--- /dev/null
+++ b/capa/features/extractors/viv/global_.py
@@ -0,0 +1,24 @@
+import logging
+
+import envi.archs.i386
+import envi.archs.amd64
+
+from capa.features.common import ARCH_I386, ARCH_AMD64, Arch
+
+logger = logging.getLogger(__name__)
+
+
+def extract_arch(vw):
+    if isinstance(vw.arch, envi.archs.amd64.Amd64Module):
+        yield Arch(ARCH_AMD64), 0x0
+
+    elif isinstance(vw.arch, envi.archs.i386.i386Module):
+        yield Arch(ARCH_I386), 0x0
+
+    else:
+        # we likely end up here:
+        #  1. handling a new architecture (e.g. aarch64)
+        #
+        # for (1), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported architecture: %s", vw.arch.__class__.__name__)
+        return
diff --git a/capa/features/extractors/viv/insn.py b/capa/features/extractors/viv/insn.py
index ffe9e3c9..5157b75b 100644
--- a/capa/features/extractors/viv/insn.py
+++ b/capa/features/extractors/viv/insn.py
@@ -19,8 +19,8 @@ import capa.features.extractors.helpers
 import capa.features.extractors.viv.helpers
 from capa.features.insn import API, Number, Offset, Mnemonic
 from capa.features.common import (
-    ARCH_X32,
-    ARCH_X64,
+    BITNESS_X32,
+    BITNESS_X64,
     MAX_BYTES_FEATURE_SIZE,
     THUNK_CHAIN_DEPTH_DELTA,
     Bytes,
@@ -34,12 +34,12 @@ from capa.features.extractors.viv.indirect_calls import NotFoundError, resolve_i
 SECURITY_COOKIE_BYTES_DELTA = 0x40
 
 
-def get_arch(vw):
-    arch = vw.getMeta("Architecture")
-    if arch == "i386":
-        return ARCH_X32
-    elif arch == "amd64":
-        return ARCH_X64
+def get_bitness(vw):
+    bitness = vw.getMeta("Architecture")
+    if bitness == "i386":
+        return BITNESS_X32
+    elif bitness == "amd64":
+        return BITNESS_X64
 
 
 def interface_extract_instruction_XXX(f, bb, insn):
@@ -193,7 +193,7 @@ def extract_insn_number_features(f, bb, insn):
             return
 
         yield Number(v), insn.va
-        yield Number(v, arch=get_arch(f.vw)), insn.va
+        yield Number(v, bitness=get_bitness(f.vw)), insn.va
 
 
 def derefs(vw, p):
@@ -389,7 +389,7 @@ def extract_insn_offset_features(f, bb, insn):
             v = oper.disp
 
             yield Offset(v), insn.va
-            yield Offset(v, arch=get_arch(f.vw)), insn.va
+            yield Offset(v, bitness=get_bitness(f.vw)), insn.va
 
         # like: [esi + ecx + 16384]
         #        reg   ^     ^
@@ -400,7 +400,7 @@ def extract_insn_offset_features(f, bb, insn):
             v = oper.disp
 
             yield Offset(v), insn.va
-            yield Offset(v, arch=get_arch(f.vw)), insn.va
+            yield Offset(v, bitness=get_bitness(f.vw)), insn.va
 
 
 def is_security_cookie(f, bb, insn) -> bool:
diff --git a/capa/features/insn.py b/capa/features/insn.py
index 6c7e07ff..c5bc727e 100644
--- a/capa/features/insn.py
+++ b/capa/features/insn.py
@@ -21,16 +21,16 @@ class API(Feature):
 
 
 class Number(Feature):
-    def __init__(self, value: int, arch=None, description=None):
-        super(Number, self).__init__(value, arch=arch, description=description)
+    def __init__(self, value: int, bitness=None, description=None):
+        super(Number, self).__init__(value, bitness=bitness, description=description)
 
     def get_value_str(self):
         return capa.render.utils.hex(self.value)
 
 
 class Offset(Feature):
-    def __init__(self, value: int, arch=None, description=None):
-        super(Offset, self).__init__(value, arch=arch, description=description)
+    def __init__(self, value: int, bitness=None, description=None):
+        super(Offset, self).__init__(value, bitness=bitness, description=description)
 
     def get_value_str(self):
         return capa.render.utils.hex(self.value)
diff --git a/capa/helpers.py b/capa/helpers.py
index dd0ff79e..1bdad425 100644
--- a/capa/helpers.py
+++ b/capa/helpers.py
@@ -21,3 +21,12 @@ def get_file_taste(sample_path: str) -> bytes:
     with open(sample_path, "rb") as f:
         taste = f.read(8)
     return taste
+
+
+def is_runtime_ida():
+    try:
+        import idc
+    except ImportError:
+        return False
+    else:
+        return True
diff --git a/capa/ida/helpers.py b/capa/ida/helpers.py
index b2e8f0fc..8cc6d894 100644
--- a/capa/ida/helpers.py
+++ b/capa/ida/helpers.py
@@ -12,6 +12,8 @@ import datetime
 import idc
 import idaapi
 import idautils
+import ida_bytes
+import ida_loader
 
 import capa
 import capa.version
@@ -28,11 +30,12 @@ SUPPORTED_IDA_VERSIONS = [
     "7.6",
 ]
 
-# file type names as returned by idaapi.get_file_type_name()
+# file type names as returned by idainfo.file_type
 SUPPORTED_FILE_TYPES = [
-    "Portable executable for 80386 (PE)",
-    "Portable executable for AMD64 (PE)",
-    "Binary file",  # x86/AMD64 shellcode support
+    idaapi.f_PE,
+    idaapi.f_ELF,
+    # idaapi.f_MACHO,
+    idaapi.f_BIN,
 ]
 
 
@@ -53,10 +56,10 @@ def is_supported_ida_version():
 
 
 def is_supported_file_type():
-    file_type = idaapi.get_file_type_name()
-    if file_type not in SUPPORTED_FILE_TYPES:
+    file_info = idaapi.get_inf_structure()
+    if file_info.filetype not in SUPPORTED_FILE_TYPES:
         logger.error("-" * 80)
-        logger.error(" Input file does not appear to be a PE file.")
+        logger.error(" Input file does not appear to be a supported file type.")
         logger.error(" ")
         logger.error(
             " capa currently only supports analyzing PE files (or binary files containing x86/AMD64 shellcode) with IDA."
@@ -121,3 +124,30 @@ def collect_metadata():
         },
         "version": capa.version.__version__,
     }
+
+
+class IDAIO:
+    """
+    An object that acts as a file-like object,
+    using bytes from the current IDB workspace.
+    """
+
+    def __init__(self):
+        super(IDAIO, self).__init__()
+        self.offset = 0
+
+    def seek(self, offset, whence=0):
+        assert whence == 0
+        self.offset = offset
+
+    def read(self, size):
+        ea = ida_loader.get_fileregion_ea(self.offset)
+        if ea == idc.BADADDR:
+            # best guess, such as if file is mapped at address 0x0.
+            ea = self.offset
+
+        logger.debug("reading 0x%x bytes at 0x%x (ea: 0x%x)", size, self.offset, ea)
+        return ida_bytes.get_bytes(ea, size)
+
+    def close(self):
+        return
diff --git a/capa/ida/plugin/view.py b/capa/ida/plugin/view.py
index 3f13b89d..3986114a 100644
--- a/capa/ida/plugin/view.py
+++ b/capa/ida/plugin/view.py
@@ -1007,7 +1007,11 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
                         self.parent_items[feature], (format_feature(feature), format_address(ea)), feature=feature
                     )
             else:
-                ea = eas.pop()
+                if eas:
+                    ea = eas.pop()
+                else:
+                    # some features may not have an address e.g. "format"
+                    ea = ""
                 for (i, v) in enumerate((format_feature(feature), format_address(ea))):
                     self.parent_items[feature].setText(i, v)
                 self.parent_items[feature].setData(0, 0x100, feature)
diff --git a/capa/main.py b/capa/main.py
index 6f6ce168..2b91fe26 100644
--- a/capa/main.py
+++ b/capa/main.py
@@ -21,7 +21,7 @@ import textwrap
 import itertools
 import contextlib
 import collections
-from typing import Any, Dict, List, Tuple, Iterable
+from typing import Any, Dict, List, Tuple
 
 import halo
 import tqdm
@@ -37,6 +37,7 @@ import capa.features.common
 import capa.features.freeze
 import capa.render.vverbose
 import capa.features.extractors
+import capa.features.extractors.common
 import capa.features.extractors.pefile
 from capa.rules import Rule, RuleSet
 from capa.engine import FeatureSet, MatchResults
@@ -45,7 +46,6 @@ from capa.features.extractors.base_extractor import FunctionHandle, FeatureExtra
 
 RULES_PATH_DEFAULT_STRING = "(embedded rules)"
 SIGNATURES_PATH_DEFAULT_STRING = "(embedded signatures)"
-SUPPORTED_FILE_MAGIC = (b"MZ", b"\x7fELF")
 BACKEND_VIV = "vivisect"
 BACKEND_SMDA = "smda"
 EXTENSIONS_SHELLCODE_32 = ("sc32", "raw32")
@@ -237,16 +237,61 @@ def has_file_limitation(rules: RuleSet, capabilities: MatchResults, is_standalon
     return False
 
 
-def is_supported_file_type(sample: str) -> bool:
+def is_supported_format(sample: str) -> bool:
     """
     Return if this is a supported file based on magic header values
     """
     with open(sample, "rb") as f:
-        magic = f.read(4)
-    if magic.startswith(SUPPORTED_FILE_MAGIC):
-        return True
-    else:
-        return False
+        taste = f.read(0x100)
+
+    return len(list(capa.features.extractors.common.extract_format(taste))) == 1
+
+
+def get_format(sample: str) -> str:
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    for feature, _ in capa.features.extractors.common.extract_format(buf):
+        assert isinstance(feature.value, str)
+        return feature.value
+
+    return "unknown"
+
+
+def is_supported_arch(sample: str) -> bool:
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    return len(list(capa.features.extractors.common.extract_arch(buf))) == 1
+
+
+def get_arch(sample: str) -> str:
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    for feature, _ in capa.features.extractors.common.extract_arch(buf):
+        assert isinstance(feature.value, str)
+        return feature.value
+
+    return "unknown"
+
+
+def is_supported_os(sample: str) -> bool:
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    return len(list(capa.features.extractors.common.extract_os(buf))) == 1
+
+
+def get_os(sample: str) -> str:
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    for feature, _ in capa.features.extractors.common.extract_os(buf):
+        assert isinstance(feature.value, str)
+        return feature.value
+
+    return "unknown"
 
 
 SHELLCODE_BASE = 0x690000
@@ -391,6 +436,14 @@ class UnsupportedFormatError(ValueError):
     pass
 
 
+class UnsupportedArchError(ValueError):
+    pass
+
+
+class UnsupportedOSError(ValueError):
+    pass
+
+
 def get_workspace(path, format, sigpaths):
     """
     load the program at the given path into a vivisect workspace using the given format.
@@ -411,7 +464,7 @@ def get_workspace(path, format, sigpaths):
 
     logger.debug("generating vivisect workspace for: %s", path)
     if format == "auto":
-        if not is_supported_file_type(path):
+        if not is_supported_format(path):
             raise UnsupportedFormatError()
 
         # don't analyze, so that we can add our Flirt function analyzer first.
@@ -443,8 +496,20 @@ def get_extractor(
 ) -> FeatureExtractor:
     """
     raises:
-      UnsupportedFormatError:
+      UnsupportedFormatError
+      UnsupportedArchError
+      UnsupportedOSError
     """
+    if format not in ("sc32", "sc64"):
+        if not is_supported_format(path):
+            raise UnsupportedFormatError()
+
+        if not is_supported_arch(path):
+            raise UnsupportedArchError()
+
+        if not is_supported_os(path):
+            raise UnsupportedOSError()
+
     if backend == "smda":
         from smda.SmdaConfig import SmdaConfig
         from smda.Disassembler import Disassembler
@@ -463,10 +528,6 @@ def get_extractor(
         import capa.features.extractors.viv.extractor
 
         with halo.Halo(text="analyzing program", spinner="simpleDots", stream=sys.stderr, enabled=not disable_progress):
-            if format == "auto" and path.endswith(EXTENSIONS_SHELLCODE_32):
-                format = "sc32"
-            elif format == "auto" and path.endswith(EXTENSIONS_SHELLCODE_64):
-                format = "sc64"
             vw = get_workspace(path, format, sigpaths)
 
             if should_save_workspace:
@@ -574,7 +635,7 @@ def get_signatures(sigs_path):
     return paths
 
 
-def collect_metadata(argv, sample_path, rules_path, format, extractor):
+def collect_metadata(argv, sample_path, rules_path, extractor):
     md5 = hashlib.md5()
     sha1 = hashlib.sha1()
     sha256 = hashlib.sha256()
@@ -589,6 +650,10 @@ def collect_metadata(argv, sample_path, rules_path, format, extractor):
     if rules_path != RULES_PATH_DEFAULT_STRING:
         rules_path = os.path.abspath(os.path.normpath(rules_path))
 
+    format = get_format(sample_path)
+    arch = get_arch(sample_path)
+    os_ = get_os(sample_path)
+
     return {
         "timestamp": datetime.datetime.now().isoformat(),
         "version": capa.version.__version__,
@@ -601,6 +666,8 @@ def collect_metadata(argv, sample_path, rules_path, format, extractor):
         },
         "analysis": {
             "format": format,
+            "arch": arch,
+            "os": os_,
             "extractor": extractor.__class__.__name__,
             "rules": rules_path,
             "base_address": extractor.get_base_address(),
@@ -909,6 +976,11 @@ def main(argv=None):
             extractor = capa.features.freeze.load(f.read())
     else:
         format = args.format
+        if format == "auto" and args.sample.endswith(EXTENSIONS_SHELLCODE_32):
+            format = "sc32"
+        elif format == "auto" and args.sample.endswith(EXTENSIONS_SHELLCODE_64):
+            format = "sc64"
+
         should_save_workspace = os.environ.get("CAPA_SAVE_WORKSPACE") not in ("0", "no", "NO", "n", None)
 
         try:
@@ -917,16 +989,32 @@ def main(argv=None):
             )
         except UnsupportedFormatError:
             logger.error("-" * 80)
-            logger.error(" Input file does not appear to be a PE file.")
+            logger.error(" Input file does not appear to be a PE or ELF file.")
             logger.error(" ")
             logger.error(
-                " capa currently only supports analyzing PE files (or shellcode, when using --format sc32|sc64)."
+                " capa currently only supports analyzing PE and ELF files (or shellcode, when using --format sc32|sc64)."
             )
             logger.error(" If you don't know the input file type, you can try using the `file` utility to guess it.")
             logger.error("-" * 80)
             return -1
+        except UnsupportedArchError:
+            logger.error("-" * 80)
+            logger.error(" Input file does not appear to target a supported architecture.")
+            logger.error(" ")
+            logger.error(" capa currently only supports analyzing x86 (32- and 64-bit).")
+            logger.error("-" * 80)
+            return -1
+        except UnsupportedOSError:
+            logger.error("-" * 80)
+            logger.error(" Input file does not appear to target a supported OS.")
+            logger.error(" ")
+            logger.error(
+                " capa currently only supports analyzing executables for some operating systems (including Windows and Linux)."
+            )
+            logger.error("-" * 80)
+            return -1
 
-    meta = collect_metadata(argv, args.sample, args.rules, format, extractor)
+    meta = collect_metadata(argv, args.sample, args.rules, extractor)
 
     capabilities, counts = find_capabilities(rules, extractor, disable_progress=args.quiet)
     meta["analysis"].update(counts)
diff --git a/capa/render/default.py b/capa/render/default.py
index 0f3f9ef3..2ed06a66 100644
--- a/capa/render/default.py
+++ b/capa/render/default.py
@@ -7,7 +7,6 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import collections
-from typing import Dict, List
 
 import tabulate
 
@@ -33,6 +32,9 @@ def render_meta(doc, ostream: StringIO):
         (width("md5", 22), width(doc["meta"]["sample"]["md5"], 82)),
         ("sha1", doc["meta"]["sample"]["sha1"]),
         ("sha256", doc["meta"]["sample"]["sha256"]),
+        ("os", doc["meta"]["analysis"]["os"]),
+        ("format", doc["meta"]["analysis"]["format"]),
+        ("arch", doc["meta"]["analysis"]["arch"]),
         ("path", doc["meta"]["sample"]["path"]),
     ]
 
diff --git a/capa/render/verbose.py b/capa/render/verbose.py
index df8a9586..d7908268 100644
--- a/capa/render/verbose.py
+++ b/capa/render/verbose.py
@@ -41,7 +41,9 @@ def render_meta(ostream, doc):
         path                 /tmp/suspicious.dll_
         timestamp            2020-07-03T10:17:05.796933
         capa version         0.0.0
-        format               auto
+        os                   windows
+        format               pe
+        arch                 amd64
         extractor            VivisectFeatureExtractor
         base address         0x10000000
         rules                (embedded rules)
@@ -55,7 +57,9 @@ def render_meta(ostream, doc):
         ("path", doc["meta"]["sample"]["path"]),
         ("timestamp", doc["meta"]["timestamp"]),
         ("capa version", doc["meta"]["version"]),
+        ("os", doc["meta"]["analysis"]["os"]),
         ("format", doc["meta"]["analysis"]["format"]),
+        ("arch", doc["meta"]["analysis"]["arch"]),
         ("extractor", doc["meta"]["analysis"]["extractor"]),
         ("base address", hex(doc["meta"]["analysis"]["base_address"])),
         ("rules", doc["meta"]["analysis"]["rules"]),
diff --git a/capa/render/vverbose.py b/capa/render/vverbose.py
index 95239eab..adecde36 100644
--- a/capa/render/vverbose.py
+++ b/capa/render/vverbose.py
@@ -113,7 +113,8 @@ def render_feature(ostream, match, feature, indent=0):
                 ostream.write(capa.rules.DESCRIPTION_SEPARATOR)
                 ostream.write(feature["description"])
 
-        render_locations(ostream, match)
+        if key not in ("os", "arch"):
+            render_locations(ostream, match)
         ostream.write("\n")
     else:
         # like:
diff --git a/capa/rules.py b/capa/rules.py
index 6fa9a4c4..b8b90faa 100644
--- a/capa/rules.py
+++ b/capa/rules.py
@@ -22,7 +22,7 @@ except ImportError:
     # https://github.com/python/mypy/issues/1153
     from backports.functools_lru_cache import lru_cache  # type: ignore
 
-from typing import Any, Set, Dict, List, Union, Iterator
+from typing import Any, Dict, List, Union, Iterator
 
 import yaml
 import ruamel.yaml
@@ -78,6 +78,9 @@ SUPPORTED_FEATURES = {
         capa.features.file.FunctionName,
         capa.features.common.Characteristic("embedded pe"),
         capa.features.common.String,
+        capa.features.common.Format,
+        capa.features.common.OS,
+        capa.features.common.Arch,
     },
     FUNCTION_SCOPE: {
         # plus basic block scope features, see below
@@ -86,6 +89,8 @@ SUPPORTED_FEATURES = {
         capa.features.common.Characteristic("calls to"),
         capa.features.common.Characteristic("loop"),
         capa.features.common.Characteristic("recursive call"),
+        capa.features.common.OS,
+        capa.features.common.Arch,
     },
     BASIC_BLOCK_SCOPE: {
         capa.features.common.MatchedRule,
@@ -103,6 +108,8 @@ SUPPORTED_FEATURES = {
         capa.features.common.Characteristic("tight loop"),
         capa.features.common.Characteristic("stack string"),
         capa.features.common.Characteristic("indirect call"),
+        capa.features.common.OS,
+        capa.features.common.Arch,
     },
 }
 
@@ -153,14 +160,14 @@ def ensure_feature_valid_for_scope(scope: str, feature: Union[Feature, Statement
         and isinstance(feature.value, str)
         and capa.features.common.Characteristic(feature.value) not in SUPPORTED_FEATURES[scope]
     ):
-        raise InvalidRule("feature %s not support for scope %s" % (feature, scope))
+        raise InvalidRule("feature %s not supported for scope %s" % (feature, scope))
 
     if not isinstance(feature, capa.features.common.Characteristic):
         # features of this scope that are not Characteristics will be Type instances.
         # check that the given feature is one of these types.
         types_for_scope = filter(lambda t: isinstance(t, type), SUPPORTED_FEATURES[scope])
         if not isinstance(feature, tuple(types_for_scope)):  # type: ignore
-            raise InvalidRule("feature %s not support for scope %s" % (feature, scope))
+            raise InvalidRule("feature %s not supported for scope %s" % (feature, scope))
 
 
 def parse_int(s: str) -> int:
@@ -217,19 +224,19 @@ def parse_feature(key: str):
     elif key == "number":
         return capa.features.insn.Number
     elif key.startswith("number/"):
-        arch = key.partition("/")[2]
+        bitness = key.partition("/")[2]
         # the other handlers here return constructors for features,
         # and we want to as well,
-        # however, we need to preconfigure one of the arguments (`arch`).
+        # however, we need to preconfigure one of the arguments (`bitness`).
         # so, instead we return a partially-applied function that
-        #  provides `arch` to the feature constructor.
+        #  provides `bitness` to the feature constructor.
         # it forwards any other arguments provided to the closure along to the constructor.
-        return functools.partial(capa.features.insn.Number, arch=arch)
+        return functools.partial(capa.features.insn.Number, bitness=bitness)
     elif key == "offset":
         return capa.features.insn.Offset
     elif key.startswith("offset/"):
-        arch = key.partition("/")[2]
-        return functools.partial(capa.features.insn.Offset, arch=arch)
+        bitness = key.partition("/")[2]
+        return functools.partial(capa.features.insn.Offset, bitness=bitness)
     elif key == "mnemonic":
         return capa.features.insn.Mnemonic
     elif key == "basic blocks":
@@ -246,6 +253,12 @@ def parse_feature(key: str):
         return capa.features.common.MatchedRule
     elif key == "function-name":
         return capa.features.file.FunctionName
+    elif key == "os":
+        return capa.features.common.OS
+    elif key == "format":
+        return capa.features.common.Format
+    elif key == "arch":
+        return capa.features.common.Arch
     else:
         raise InvalidRule("unexpected statement: %s" % key)
 
diff --git a/scripts/bulk-process.py b/scripts/bulk-process.py
index da80a477..3bb3495e 100644
--- a/scripts/bulk-process.py
+++ b/scripts/bulk-process.py
@@ -126,7 +126,7 @@ def get_capa_results(args):
             "error": "unexpected error: %s" % (e),
         }
 
-    meta = capa.main.collect_metadata("", path, "", format, extractor)
+    meta = capa.main.collect_metadata("", path, "", extractor)
     capabilities, counts = capa.main.find_capabilities(rules, extractor, disable_progress=True)
     meta["analysis"].update(counts)
 
diff --git a/scripts/capa2yara.py b/scripts/capa2yara.py
index 169b8cc4..5a72c697 100644
--- a/scripts/capa2yara.py
+++ b/scripts/capa2yara.py
@@ -43,7 +43,7 @@ import capa.rules
 import capa.engine
 import capa.features
 import capa.features.insn
-from capa.features.common import ARCH_X32, ARCH_X64, String
+from capa.features.common import BITNESS_X32, BITNESS_X64, String
 
 logger = logging.getLogger("capa2yara")
 
diff --git a/scripts/capa_as_library.py b/scripts/capa_as_library.py
index 449c35b9..012c1bf3 100644
--- a/scripts/capa_as_library.py
+++ b/scripts/capa_as_library.py
@@ -169,7 +169,7 @@ def capa_details(file_path, output_format="dictionary"):
     capabilities, counts = capa.main.find_capabilities(rules, extractor, disable_progress=True)
 
     # collect metadata (used only to make rendering more complete)
-    meta = capa.main.collect_metadata("", file_path, RULES_PATH, "auto", extractor)
+    meta = capa.main.collect_metadata("", file_path, RULES_PATH, extractor)
     meta["analysis"].update(counts)
 
     capa_output = False
diff --git a/scripts/detect-elf-os.py b/scripts/detect-elf-os.py
new file mode 100644
index 00000000..c01990f4
--- /dev/null
+++ b/scripts/detect-elf-os.py
@@ -0,0 +1,74 @@
+#!/usr/bin/env python2
+"""
+Copyright (C) 2021 FireEye, Inc. All Rights Reserved.
+Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+You may obtain a copy of the License at: [package root]/LICENSE.txt
+Unless required by applicable law or agreed to in writing, software distributed under the License
+ is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and limitations under the License.
+
+detect-elf-os
+
+Attempt to detect the underlying OS that the given ELF file targets.
+"""
+import sys
+import logging
+import argparse
+import contextlib
+from typing import BinaryIO
+
+import capa.helpers
+import capa.features.extractors.elf
+
+logger = logging.getLogger("capa.detect-elf-os")
+
+
+def main(argv=None):
+    if capa.helpers.is_runtime_ida():
+        from capa.ida.helpers import IDAIO
+
+        f: BinaryIO = IDAIO()
+
+    else:
+        if argv is None:
+            argv = sys.argv[1:]
+
+        parser = argparse.ArgumentParser(description="Detect the underlying OS for the given ELF file")
+        parser.add_argument("sample", type=str, help="path to ELF file")
+
+        logging_group = parser.add_argument_group("logging arguments")
+
+        logging_group.add_argument("-d", "--debug", action="store_true", help="enable debugging output on STDERR")
+        logging_group.add_argument(
+            "-q", "--quiet", action="store_true", help="disable all status output except fatal errors"
+        )
+
+        args = parser.parse_args(args=argv)
+
+        if args.quiet:
+            logging.basicConfig(level=logging.WARNING)
+            logging.getLogger().setLevel(logging.WARNING)
+        elif args.debug:
+            logging.basicConfig(level=logging.DEBUG)
+            logging.getLogger().setLevel(logging.DEBUG)
+        else:
+            logging.basicConfig(level=logging.INFO)
+            logging.getLogger().setLevel(logging.INFO)
+
+        f = open(args.sample, "rb")
+
+    with contextlib.closing(f):
+        try:
+            print(capa.features.extractors.elf.detect_elf_os(f))
+            return 0
+        except capa.features.extractors.elf.CorruptElfFile as e:
+            logger.error("corrupt ELF file: %s", str(e.args[0]))
+            return -1
+
+
+if __name__ == "__main__":
+    if capa.helpers.is_runtime_ida():
+        main()
+    else:
+        sys.exit(main())
diff --git a/scripts/show-capabilities-by-function.py b/scripts/show-capabilities-by-function.py
index b95481df..11fcdaab 100644
--- a/scripts/show-capabilities-by-function.py
+++ b/scripts/show-capabilities-by-function.py
@@ -171,7 +171,7 @@ def main(argv=None):
             logger.error("-" * 80)
             return -1
 
-    meta = capa.main.collect_metadata(argv, args.sample, args.rules, format, extractor)
+    meta = capa.main.collect_metadata(argv, args.sample, args.rules, extractor)
     capabilities, counts = capa.main.find_capabilities(rules, extractor)
     meta["analysis"].update(counts)
 
diff --git a/scripts/show-features.py b/scripts/show-features.py
index 3090a471..6d69ea7f 100644
--- a/scripts/show-features.py
+++ b/scripts/show-features.py
@@ -75,6 +75,7 @@ import capa.rules
 import capa.engine
 import capa.helpers
 import capa.features
+import capa.features.common
 import capa.features.freeze
 
 logger = logging.getLogger("capa.show-features")
@@ -203,14 +204,23 @@ def print_features(functions, extractor):
             continue
 
         for feature, va in extractor.extract_function_features(f):
+            if capa.features.common.is_global_feature(feature):
+                continue
+
             print("func: 0x%08x: %s" % (va, feature))
 
         for bb in extractor.get_basic_blocks(f):
             for feature, va in extractor.extract_basic_block_features(f, bb):
+                if capa.features.common.is_global_feature(feature):
+                    continue
+
                 print("bb  : 0x%08x: %s" % (va, feature))
 
             for insn in extractor.get_instructions(f, bb):
                 for feature, va in extractor.extract_insn_features(f, bb, insn):
+                    if capa.features.common.is_global_feature(feature):
+                        continue
+
                     try:
                         print("insn: 0x%08x: %s" % (va, feature))
                     except UnicodeEncodeError:
diff --git a/tests/fixtures.py b/tests/fixtures.py
index f15e7f06..b47e79e2 100644
--- a/tests/fixtures.py
+++ b/tests/fixtures.py
@@ -21,7 +21,19 @@ import capa.features.file
 import capa.features.insn
 import capa.features.common
 import capa.features.basicblock
-from capa.features.common import ARCH_X32, ARCH_X64
+from capa.features.common import (
+    OS,
+    OS_LINUX,
+    ARCH_I386,
+    FORMAT_PE,
+    ARCH_AMD64,
+    FORMAT_ELF,
+    OS_WINDOWS,
+    BITNESS_X32,
+    BITNESS_X64,
+    Arch,
+    Format,
+)
 
 CD = os.path.dirname(__file__)
 
@@ -196,6 +208,8 @@ def get_data_path_by_name(name):
         return os.path.join(CD, "data", "773290480d5445f11d3dc1b800728966.exe_")
     elif name.startswith("3b13b"):
         return os.path.join(CD, "data", "3b13b6f1d7cd14dc4a097a12e2e505c0a4cff495262261e2bfc991df238b9b04.dll_")
+    elif name == "7351f.elf":
+        return os.path.join(CD, "data", "7351f8a40c5450557b24622417fc478d.elf_")
     else:
         raise ValueError("unexpected sample fixture: %s" % name)
 
@@ -241,6 +255,8 @@ def get_sample_md5_by_name(name):
     elif name.startswith("3b13b"):
         # file name is SHA256 hash
         return "56a6ffe6a02941028cc8235204eef31d"
+    elif name == "7351f.elf":
+        return "7351f8a40c5450557b24622417fc478d"
     else:
         raise ValueError("unexpected sample fixture: %s" % name)
 
@@ -379,10 +395,10 @@ FEATURE_PRESENCE_TESTS = sorted(
         # insn/number: stack adjustments
         ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xC), False),
         ("mimikatz", "function=0x40105D", capa.features.insn.Number(0x10), False),
-        # insn/number: arch flavors
+        # insn/number: bitness flavors
         ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xFF), True),
-        ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xFF, arch=ARCH_X32), True),
-        ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xFF, arch=ARCH_X64), False),
+        ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xFF, bitness=BITNESS_X32), True),
+        ("mimikatz", "function=0x40105D", capa.features.insn.Number(0xFF, bitness=BITNESS_X64), False),
         # insn/offset
         ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0), True),
         ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x4), True),
@@ -395,10 +411,10 @@ FEATURE_PRESENCE_TESTS = sorted(
         # insn/offset: negative
         ("mimikatz", "function=0x4011FB", capa.features.insn.Offset(-0x1), True),
         ("mimikatz", "function=0x4011FB", capa.features.insn.Offset(-0x2), True),
-        # insn/offset: arch flavors
+        # insn/offset: bitness flavors
         ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0), True),
-        ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0, arch=ARCH_X32), True),
-        ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0, arch=ARCH_X64), False),
+        ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0, bitness=BITNESS_X32), True),
+        ("mimikatz", "function=0x40105D", capa.features.insn.Offset(0x0, bitness=BITNESS_X64), False),
         # insn/api
         ("mimikatz", "function=0x403BAC", capa.features.insn.API("advapi32.CryptAcquireContextW"), True),
         ("mimikatz", "function=0x403BAC", capa.features.insn.API("advapi32.CryptAcquireContext"), True),
@@ -499,6 +515,26 @@ FEATURE_PRESENCE_TESTS = sorted(
         ("mimikatz", "function=0x456BB9", capa.features.common.Characteristic("calls to"), False),
         # file/function-name
         ("pma16-01", "file", capa.features.file.FunctionName("__aulldiv"), True),
+        # os & format & arch
+        ("pma16-01", "file", OS(OS_WINDOWS), True),
+        ("pma16-01", "file", OS(OS_LINUX), False),
+        ("pma16-01", "function=0x404356", OS(OS_WINDOWS), True),
+        ("pma16-01", "function=0x404356,bb=0x4043B9", OS(OS_WINDOWS), True),
+        ("pma16-01", "file", Arch(ARCH_I386), True),
+        ("pma16-01", "file", Arch(ARCH_AMD64), False),
+        ("pma16-01", "function=0x404356", Arch(ARCH_I386), True),
+        ("pma16-01", "function=0x404356,bb=0x4043B9", Arch(ARCH_I386), True),
+        ("pma16-01", "file", Format(FORMAT_PE), True),
+        ("pma16-01", "file", Format(FORMAT_ELF), False),
+        # elf support
+        ("7351f.elf", "file", OS(OS_LINUX), True),
+        ("7351f.elf", "file", OS(OS_WINDOWS), False),
+        ("7351f.elf", "file", Format(FORMAT_ELF), True),
+        ("7351f.elf", "file", Format(FORMAT_PE), False),
+        ("7351f.elf", "file", Arch(ARCH_I386), False),
+        ("7351f.elf", "file", Arch(ARCH_AMD64), True),
+        ("7351f.elf", "function=0x408753", capa.features.common.String("/dev/null"), True),
+        ("7351f.elf", "function=0x408753,bb=0x408781", capa.features.insn.API("open"), True),
     ],
     # order tests by (file, item)
     # so that our LRU cache is most effective.
diff --git a/tests/test_engine.py b/tests/test_engine.py
index 642ddb49..57fffb8e 100644
--- a/tests/test_engine.py
+++ b/tests/test_engine.py
@@ -474,11 +474,11 @@ def test_match_namespace():
 
 def test_render_number():
     assert str(capa.features.insn.Number(1)) == "number(0x1)"
-    assert str(capa.features.insn.Number(1, arch=capa.features.common.ARCH_X32)) == "number/x32(0x1)"
-    assert str(capa.features.insn.Number(1, arch=capa.features.common.ARCH_X64)) == "number/x64(0x1)"
+    assert str(capa.features.insn.Number(1, bitness=capa.features.common.BITNESS_X32)) == "number/x32(0x1)"
+    assert str(capa.features.insn.Number(1, bitness=capa.features.common.BITNESS_X64)) == "number/x64(0x1)"
 
 
 def test_render_offset():
     assert str(capa.features.insn.Offset(1)) == "offset(0x1)"
-    assert str(capa.features.insn.Offset(1, arch=capa.features.common.ARCH_X32)) == "offset/x32(0x1)"
-    assert str(capa.features.insn.Offset(1, arch=capa.features.common.ARCH_X64)) == "offset/x64(0x1)"
+    assert str(capa.features.insn.Offset(1, bitness=capa.features.common.BITNESS_X32)) == "offset/x32(0x1)"
+    assert str(capa.features.insn.Offset(1, bitness=capa.features.common.BITNESS_X64)) == "offset/x64(0x1)"
diff --git a/tests/test_pefile_features.py b/tests/test_pefile_features.py
index 8bb46d43..2e1afc7b 100644
--- a/tests/test_pefile_features.py
+++ b/tests/test_pefile_features.py
@@ -20,9 +20,12 @@ import capa.features.file
 )
 def test_pefile_features(sample, scope, feature, expected):
     if scope.__name__ != "file":
-        pytest.xfail("pefile only extract file scope features")
+        pytest.xfail("pefile only extracts file scope features")
 
     if isinstance(feature, capa.features.file.FunctionName):
-        pytest.xfail("pefile only doesn't extract function names")
+        pytest.xfail("pefile doesn't extract function names")
+
+    if ".elf" in sample:
+        pytest.xfail("pefile doesn't handle ELF files")
 
     fixtures.do_test_feature_presence(fixtures.get_pefile_extractor, sample, scope, feature, expected)
diff --git a/tests/test_rules.py b/tests/test_rules.py
index 12791d2e..9362ef91 100644
--- a/tests/test_rules.py
+++ b/tests/test_rules.py
@@ -15,7 +15,20 @@ import capa.engine
 import capa.features.common
 from capa.features.file import FunctionName
 from capa.features.insn import Number, Offset
-from capa.features.common import ARCH_X32, ARCH_X64, String
+from capa.features.common import (
+    OS,
+    OS_LINUX,
+    ARCH_I386,
+    FORMAT_PE,
+    ARCH_AMD64,
+    FORMAT_ELF,
+    OS_WINDOWS,
+    BITNESS_X32,
+    BITNESS_X64,
+    Arch,
+    Format,
+    String,
+)
 
 
 def test_rule_ctor():
@@ -517,7 +530,7 @@ def test_invalid_number():
         )
 
 
-def test_number_arch():
+def test_number_bitness():
     r = capa.rules.Rule.from_yaml(
         textwrap.dedent(
             """
@@ -529,13 +542,13 @@ def test_number_arch():
             """
         )
     )
-    assert r.evaluate({Number(2, arch=ARCH_X32): {1}}) == True
+    assert r.evaluate({Number(2, bitness=BITNESS_X32): {1}}) == True
 
     assert r.evaluate({Number(2): {1}}) == False
-    assert r.evaluate({Number(2, arch=ARCH_X64): {1}}) == False
+    assert r.evaluate({Number(2, bitness=BITNESS_X64): {1}}) == False
 
 
-def test_number_arch_symbol():
+def test_number_bitness_symbol():
     r = capa.rules.Rule.from_yaml(
         textwrap.dedent(
             """
@@ -547,7 +560,7 @@ def test_number_arch_symbol():
             """
         )
     )
-    assert r.evaluate({Number(2, arch=ARCH_X32, description="some constant"): {1}}) == True
+    assert r.evaluate({Number(2, bitness=BITNESS_X32, description="some constant"): {1}}) == True
 
 
 def test_offset_symbol():
@@ -595,7 +608,7 @@ def test_count_offset_symbol():
     assert r.evaluate({Offset(0x100, description="symbol name"): {1, 2, 3}}) == True
 
 
-def test_offset_arch():
+def test_offset_bitness():
     r = capa.rules.Rule.from_yaml(
         textwrap.dedent(
             """
@@ -607,13 +620,13 @@ def test_offset_arch():
             """
         )
     )
-    assert r.evaluate({Offset(2, arch=ARCH_X32): {1}}) == True
+    assert r.evaluate({Offset(2, bitness=BITNESS_X32): {1}}) == True
 
     assert r.evaluate({Offset(2): {1}}) == False
-    assert r.evaluate({Offset(2, arch=ARCH_X64): {1}}) == False
+    assert r.evaluate({Offset(2, bitness=BITNESS_X64): {1}}) == False
 
 
-def test_offset_arch_symbol():
+def test_offset_bitness_symbol():
     r = capa.rules.Rule.from_yaml(
         textwrap.dedent(
             """
@@ -625,7 +638,7 @@ def test_offset_arch_symbol():
             """
         )
     )
-    assert r.evaluate({Offset(2, arch=ARCH_X32, description="some constant"): {1}}) == True
+    assert r.evaluate({Offset(2, bitness=BITNESS_X32, description="some constant"): {1}}) == True
 
 
 def test_invalid_offset():
@@ -944,3 +957,57 @@ def test_function_name_features():
     assert (FunctionName("strcpy") in children) == True
     assert (FunctionName("strcmp", description="copy from here to there") in children) == True
     assert (FunctionName("strdup", description="duplicate a string") in children) == True
+
+
+def test_os_features():
+    rule = textwrap.dedent(
+        """
+        rule:
+            meta:
+                name: test rule
+                scope: file
+            features:
+                - and:
+                    - os: windows
+        """
+    )
+    r = capa.rules.Rule.from_yaml(rule)
+    children = list(r.statement.get_children())
+    assert (OS(OS_WINDOWS) in children) == True
+    assert (OS(OS_LINUX) not in children) == True
+
+
+def test_format_features():
+    rule = textwrap.dedent(
+        """
+        rule:
+            meta:
+                name: test rule
+                scope: file
+            features:
+                - and:
+                    - format: pe
+        """
+    )
+    r = capa.rules.Rule.from_yaml(rule)
+    children = list(r.statement.get_children())
+    assert (Format(FORMAT_PE) in children) == True
+    assert (Format(FORMAT_ELF) not in children) == True
+
+
+def test_arch_features():
+    rule = textwrap.dedent(
+        """
+        rule:
+            meta:
+                name: test rule
+                scope: file
+            features:
+                - and:
+                    - arch: amd64
+        """
+    )
+    r = capa.rules.Rule.from_yaml(rule)
+    children = list(r.statement.get_children())
+    assert (Arch(ARCH_AMD64) in children) == True
+    assert (Arch(ARCH_I386) not in children) == True
diff --git a/tests/test_viv_features.py b/tests/test_viv_features.py
index fa8bfda3..3989db81 100644
--- a/tests/test_viv_features.py
+++ b/tests/test_viv_features.py
@@ -15,6 +15,9 @@ from fixtures import *
     indirect=["sample", "scope"],
 )
 def test_viv_features(sample, scope, feature, expected):
+    if "elf" in sample:
+        pytest.xfail("viv ELF parsing is broken (for our test file), see #735")
+
     fixtures.do_test_feature_presence(fixtures.get_viv_extractor, sample, scope, feature, expected)
 
 
@@ -24,4 +27,7 @@ def test_viv_features(sample, scope, feature, expected):
     indirect=["sample", "scope"],
 )
 def test_viv_feature_counts(sample, scope, feature, expected):
+    if "elf" in sample:
+        pytest.xfail("viv ELF parsing is broken (for our test file), see #735")
+
     fixtures.do_test_feature_count(fixtures.get_viv_extractor, sample, scope, feature, expected)