Skip to content

lydell/userscript-proxy

 
 

Repository files navigation

Userscript Proxy

Browser extensions on iOS, Android and pretty much any other web browsing device. No jailbreak/root required.

Userscript Proxy is built around mitmproxy and acts as a MITM, injecting matching userscripts into web pages as they flow through it. Both HTTP and HTTPS are supported.

Security

Userscript Proxy can (and must be able to) read and modify all HTTP(S) traffic sent to and from the device in question, so the only reasonably secure way to use it is to run it on a server controlled by oneself.

Ignoring hosts

Apps like App Store and Facebook Messenger refuse to connect through a MITM proxy, so their traffic must be ignored by mitmproxy. There are two approaches:

  • Blacklisting hosts that cannot connect through the proxy. Tedious, because you have to add exceptions for apps and such all the time.
  • Whitelisting hosts where userscripts should be applied. Works well in general, but does not allow universal userscripts that run on all sites, and the whitelist must be updated when a new userscript is added.

Blacklisting or whitelisting is done by giving the --ignore or --intercept flag together with one or more files containing ignore/intercept rules. Examples:

# Take ignore rules from ignore.txt (included):
python3.6 launcher.py --ignore "ignore.txt"

# Take intercept rules from all .txt files whose names start with "foo":
python3.6 launcher.py --intercept "foo*.txt"

Rules can be specified in two ways:

Basic pattern

Based on the syntax used by userscript @include directives. Asterisk (*) means any string (including the empty string). *. is automatically prepended. :* is automatically appended unless the rule contains a colon (:).

To match a domain without matching all of its subdomains, use a regex rule instead (see below).

Examples

Rule Matches
site.com site.com and x.site.com
api.site.com api.site.com and x.api.site.com, but not www.site.com
*cdn.net cdn.net, fbcdn.net and x.fbcdn.net, but not cdn.net.com
site.com:80 site.com:80 and x.site.com:80, but not site.com:443

Regular expression

If a rule starts and ends with a slash (/), it is treated as a Python regex.

Note that the string to match against contains both a host and a port, e.g. example.com:443, and that the regex is used verbatim (i.e. you have to explicitly provide ^ etc if desired). The only exception is that the case-insensitivity flag (?i) is automatically added.

Also, be careful with $: A regex like /site.com$/ will never match, because it will only be used to check strings like site.com:80.

Anything from a # until the end of the line is treated as a comment. Leading and trailing whitespace have no effect.

Examples

Rule Matches
/cdn\./ fbcdn.net, cdn.site.com, cdn.x.site.com, but not cdna.com
/^site\.com:/ site.com, but not x.site.com, mysite.com or site.com.net

Data usage

Userscript Proxy has no data usage impact when no userscript is injected, i.e. for URLs without any matching userscript. When a script is injected, exactly one of the following things happens:

  • The entire userscript is injected as inline JavaScript (potentially dozens or even hundreds of kilobytes).
  • A <script> tag referencing the userscript is injected (~100 bytes, depending on the length of the @downloadURL).

A userscript is injected by reference if and only if it has a specified @downloadURL in its metadata. (This can be overridden using the --inline flag, in which case all userscripts are injected inline.)

Userscripts are injected into every response from a matching URL, and the size of a userscript can be anything from a few hundred bytes for the most basic ones to hundreds of kilobytes in extreme cases, so there are massive data usage reductions to be gained from making the userscript accessible by URL and including a @downloadURL. If the @downloadURL approach is not possible, for one reason or the other, it is a good idea to be aware of this issue, and to take appropriate action such as minifying userscripts and adding suitable ignore rules.

Userscript compatibility

Userscript Proxy supports (a subset of) the Greasemonkey metadata syntax. No adaptation of plain JavaScript userscripts should be required. These directives are supported:

  • @name
  • @version
  • @run-at document_(start|end|idle)
  • @match
  • @include (basic pattern and regex)
  • @exclude
  • @noframes
  • @downloadURL

The GM API and similar runtime facilities are not supported, because userscripts can only be injected as regular scripts.

Options

--ignore FILE/--intercept FILE

Take ignore or intercept rules from FILE, which can be a glob pattern matching multiple files. --ignore and --intercept cannot be used together. See examples above.

--inline, -i

Always inject scripts inline (<script>...</script>), never linked (<script src="..."></script>). Useful to test new userscript features without having to re-upload the userscript and clear browser cache.

--list-injected, -l

Insert an HTML comment in each page specifying which userscripts (if any) were injected.

--port PORT, -p PORT

Make mitmproxy listen to TCP port PORT. Defaults to 8080.

--query-param-to-disable PARAM, -q PARAM

Disable userscripts when the request URL contains PARAM as a query parameter. For example, use -q foo to disable userscripts for http://example.com?foo. Defaults to nouserscripts.

--recursive, -r

Recurse into directories when looking for userscripts.

--transparent, -t

Run mitmproxy in transparent mode. Useful if you cannot set a proxy in the client, e.g. when using OpenVPN Connect on Android to connect to a VPN server on the network where your proxy is running. In such cases, you have to route traffic from the client to the proxy at the network layer instead, making transparent mode necessary.

--userscripts DIR, -u DIR

Load userscripts from directory DIR. Defaults to userscripts.

About

Browser extensions on any device

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 96.6%
  • Nix 1.4%
  • JavaScript 1.3%
  • Shell 0.7%