Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
403 lines (265 sloc) 16.4 KB

Plugins

Introduction

CTFd features a plugin interface allowing for the modification of CTFd behavior without modifying the core CTFd code. This has a number of benefits over forking and modifying CTFd:

  • Your modifications and plugins can be shared more easily
  • CTFd can be updated without losing any custom behavior

The CTFd developers will do their best to not introduce breaking changes but keep in mind that the plugin interface is still under development and could change.

Tip

Official CTFd plugins are available at https://ctfd.io/store. Contact us regarding custom plugins and special projects.

Tip

Community plugins are available at https://github.com/CTFd/plugins.

Architecture

CTFd plugins are implemented as Python modules with some CTFd specific files.

CTFd
└── plugins
   └── CTFd-plugin
       ├── README.md          # README file
       ├── __init__.py        # Main code file loaded by CTFd
       ├── requirements.txt   # Any requirements that need to be installed
       └── config.json        # Plugin configuration file

Effectively CTFd will look at every folder in the CTFd/plugins folder for the load() function.

If the load() function is found, CTFd will call that function with itself (as a Flask app) as a parameter (i.e. load(app)). This is done after CTFd has added all of its internal routes but before CTFd has fully instantiated itself. This allows plugins to modify many aspects of CTFd without having to modify CTFd itself.

config.json

config.json exists to give plugin developers a way to define attributes about their plugin. It's primary usage within CTFd is to give users a way to access a Configuration or Settings page for the plugin.

This is an example config.json file:

{
    "name": "CTFd Plugin",
    "route": "/admin/custom_plugin_route"
}

This is ultimately rendered to the user with the following template snippet:

{% if plugins %}
<li>
    <a href="#" class="dropdown-toggle" data-toggle="dropdown" role="button" aria-haspopup="true" aria-expanded="false">Plugins <span class="caret"></span></a>
    <ul class="dropdown-menu">
            {% for plugin in plugins %}
                    <li><a href="{{ request.script_root }}{{ plugin.route }}">{{ plugin.name }}</a></li>
            {% endfor %}
    </ul>
</li>
{% endif %}

config.html

In the past CTFd used a static file known as config.html which existed to give plugin developers a page that is loaded by the CTFd admin panel. This has been superceded in favor of config.json but is still supported for backwards compatability.

The config.html file for a plugin is available by CTFd admins at /admin/plugins/<plugin-folder-name>. Thus if config.html is stored in CTFd-S3-plugin, it would be available at /admin/plugins/CTFd-S3-plugin.

config.html is loaded as a Jinja template so it has access to all of the same functions and abilities that CTFd exposes to Jinja. Jinja templates are technically also capable of running arbitrary Python code but this is ancillary.

Adding New Routes

Adding new routes in CTFd is effectively just an exercise in writing new Flask routes. Since the plugin itself is passed the entire app, the plugin can leverage the app.route decorator to add new routes.

A simple example is as follows:

from flask import render_template


def load(app):
    @app.route('/faq', methods=['GET'])
    def view_faq():
        return render_template('page.html', content="<h1>FAQ Page</h1>")

Modifying Existing Routes

It is slightly more complicated to override existing routes in CTFd/Flask because it is not strictly supported by Flask. The approach currently used is to modify the app.view_functions dictionary which contains the mapping of routes to the functions used to handle them.

from flask import render_template

from CTFd.models import db
from CTFd.utils import admins_only, is_admin

from CTFd import utils

def load(app):
    def view_challenges():
        return render_template('page.html', content="<h1>Challenges are currently closed</h1>")

    # The format used by the view_functions dictionary is blueprint.view_function_name
    app.view_functions['challenges.challenges_view'] = view_challenges

If for some reason you wish to add a new method to an existing route you can modify the url_map as follows:

from werkzeug.routing import Rule

app.url_map.add(Rule('/challenges', endpoint='challenges.challenges_view', methods=['GET', 'POST']))

Adding Database Tables

Sometimes CTFd doesn't have enough database tables or columns to let you do what you need. In this case you can use a plugin to create a new table and then use the information in the previous two sections to create routes or modify existing routes to access your new table.

from CTFd.models import db


class Avatars(db.Model):
    id = db.Column(db.Integer, primary_key=True)
    team = db.Column(db.Integer, db.ForeignKey('teams.id'))
    location = db.Column(db.Text)

    def __init__(self, team, location):
        self.target = team
        self.location = location


def load(app):
    app.db.create_all()
    @app.route('/profile/avatar', methods=['GET', 'POST'])
    def profile_avatars():
        raise NotImplementedError

Replacing Templates

In some situations it might make sense for your plugin to replace the logic for a single page template instead of creating an entire theme.

The utils.override_template() function allows a plugin to replace the content of a single template within CTFd such that CTFd will use the new content instead of the content in the original file.

from CTFd.utils import override_template
import os

def load(app):
    dir_path = os.path.dirname(os.path.realpath(__file__))
    template_path = os.path.join(dir_path, 'new-scoreboard.html')
    override_template('scoreboard.html', open(template_path).read())

With this code CTFd will use new-scoreboard.html instead of the scoreboard.html file it normally would have used.

Registering Assets

Very often you will want to provide users with static assets (e.g. JS, CSS). Instead of registering handlers for them on your own, you can use the CTFd built in plugin utilities, register_plugin_assets_directory and register_plugin_asset.

For example to register an entire assets directory as available to the user:

from CTFd.plugins import register_plugin_assets_directory

def load(app):
    # Available at http://ctfd/plugins/test_plugin/assets/
    register_plugin_assets_directory(app, base_path='/plugins/test_plugin/assets/')

Or to only provide a single file:

from CTFd.plugins import register_plugin_asset

def load(app):
    # Available at http://ctfd/plugins/test_plugin/assets/file.js
    register_plugin_asset(app, asset_path='/plugins/test_plugin/assets/file.js')

Challenge Types

In CTFd, there is a concept of a type of challenge. Most CTFs only ever provide challenges as a snippet of text alongside some files. CTFd expands upon this and allows developers to create new challenge types which diversify what users will see.

Ultimately, users will still read some text, and submit some value but CTFd allows you to style and customize this so users can submit data in new ways.

For example, instead of an input to submit a single flag value, you might require teams to submit multiple flags or you might create some kind of customized UI where teams need to arrange blocks or text in some order.

The approach used by CTFd here is to give each "type" of challenge an ID and a name.

Tip

You can see how CTFd implements its default standard challenge here. You can also see how CTFd implements dynamic scoring using this feature.

Each challenge is implemented as a child class of the BaseChallenge and implements static methods named create, read, update, delete, attempt, solve, and fail.

When a user attempts to solve a challenge, CTFd will look up the challenge type and then call the solve method as shown in the following snippet of code:

chal_class = get_chal_class(chal.type)
status, message = chal_class.attempt(chal, request)

if status:  # The challenge plugin says the input is right
    if ctftime() or is_admin():
        chal_class.solve(team=team, chal=chal, request=request)
    return jsonify({'status': 1, 'message': message})

else:  # The challenge plugin says the input is wrong
    if ctftime() or is_admin():
        chal_class.fail(team=team, chal=chal, request=request)

This structure allows each Challenge Type to dictate how they are attempted, solved, and marked incorrect.

The Challenge Type also dictates the database table that it uses to store data. By default this uses the type column as a polymorphic_identity to implement table inheritance. Effectively each child table will use the Challenges table as a parent. The child table can add whatever columns it wishes but still leverage the existing columns from the parent.

We can see in the following code that the polymorphic_identity is specified to be dynamic as well as the type parameter. We can also see the call to create_all() which will create the table in our database.

class DynamicChallenge(Challenges):
    __mapper_args__ = {'polymorphic_identity': 'dynamic'}
    id = db.Column(None, db.ForeignKey('challenges.id'), primary_key=True)
    initial = db.Column(db.Integer)
    minimum = db.Column(db.Integer)
    decay = db.Column(db.Integer)

    def __init__(self, name, description, value, category, type='dynamic', minimum=1, decay=50):
        self.name = name
        self.description = description
        self.value = value
        self.initial = value
        self.category = category
        self.type = type
        self.minimum = minimum
        self.decay = decay


def load(app):
    app.db.create_all()
    CHALLENGE_CLASSES['dynamic'] = DynamicValueChallenge
    register_plugin_assets_directory(app, base_path='/plugins/DynamicValueChallenge/assets/')

This code creates the necessary tables for the Challenge Type plugin which should be used in addition to the staticmethods used to define the challenge's behavior.

Every challenge type must be added to the global dictionary that specifies all challenge types:

CHALLENGE_CLASSES = {
    "standard": CTFdStandardChallenge
}


def get_chal_class(class_id):
    cls = CHALLENGE_CLASSES.get(class_id)
    if cls is None:
        raise KeyError
    return cls

The Standard Challenge type provided within CTFd can be used as a base from which to build additional Challenge Type plugins.

Once new challenges are registered, CTFd will provide a dropdown allowing you to choose from all the challenge types you can create.

Each Challenge Type contains templates and scripts dictionaries which contain the routes for HTML and JS files needed for the operation of the modals used to create and update the challenges.

These routes are not automatically defined by CTFd.

Each challenge type plugin specifies the location of their own templates and scripts. An example is the built in standard challenge type plugin. It specifies the URLs that the assets are located at for the user's browser to load:

templates = {  # Templates used for each aspect of challenge editing & viewing
    'create': '/plugins/challenges/assets/create.html',
    'update': '/plugins/challenges/assets/update.html',
    'view': '/plugins/challenges/assets/view.html',
}
scripts = {  # Scripts that are loaded when a template is loaded
    'create': '/plugins/challenges/assets/create.js',
    'update': '/plugins/challenges/assets/update.js',
    'view': '/plugins/challenges/assets/view.js',
}

These files are registered with Flask with the following code:

from CTFd.plugins import register_plugin_assets_directory

def load(app):
    register_plugin_assets_directory(app, base_path='/plugins/challenges/assets/')

The aforementioned code handles the Python logic around new challenges but in order to fully integrate with CTFd you will need to create new Nunjucks templates to give admins/teams the ability to modify/update/solve your challenge. The templates used by the Standard Challenge Type should serve as examples.

Flag Types

Flag types conversely are used to give developers a way to allow teams to submit flags which do not conform to a hardcoded string or a regex-able value.

The approach is very similar to Challenges with a base Flag/Key class and a global dictionary specifying all the Flag/Key types:

class BaseFlag(object):
    name = None
    templates = {}

    @staticmethod
    def compare(self, saved, provided):
        return True


class CTFdStaticFlag(BaseFlag):
    name = "static"
    templates = {  # Nunjucks templates used for key editing & viewing
        "create": "/plugins/flags/assets/static/create.html",
        "update": "/plugins/flags/assets/static/edit.html",
    }

    @staticmethod
    def compare(chal_key_obj, provided):
        saved = chal_key_obj.content
        data = chal_key_obj.data

        if len(saved) != len(provided):
            return False
        result = 0

        if data == "case_insensitive":
            for x, y in zip(saved.lower(), provided.lower()):
                result |= ord(x) ^ ord(y)
        else:
            for x, y in zip(saved, provided):
                result |= ord(x) ^ ord(y)
        return result == 0


class CTFdRegexFlag(BaseFlag):
    name = "regex"
    templates = {  # Nunjucks templates used for key editing & viewing
        "create": "/plugins/flags/assets/regex/create.html",
        "update": "/plugins/flags/assets/regex/edit.html",
    }

    @staticmethod
    def compare(chal_key_obj, provided):
        saved = chal_key_obj.content
        data = chal_key_obj.data

        if data == "case_insensitive":
            res = re.match(saved, provided, re.IGNORECASE)
        else:
            res = re.match(saved, provided)

        return res and res.group() == provided


FLAG_CLASSES = {"static": CTFdStaticFlag, "regex": CTFdRegexFlag}


def get_flag_class(class_id):
    cls = FLAG_CLASSES.get(class_id)
    if cls is None:
        raise KeyError
    return cls

When a challenge solution is submitted, the challenge plugin itself is responsible for:

  1. Loading the appropriate Key class using the get_flag_class() function.
  2. Properly calling the static compare() method defined by each Flag class.
  3. Returning the correctness boolean and the message displayed to the user.

This is properly implemented by the following code copied from the default standard challenge:

@staticmethod
def attempt(challenge, request):
    data = request.form or request.get_json()
    submission = data['submission'].strip()
    flags = Flags.query.filter_by(challenge_id=challenge.id).all()
    for flag in flags:
        if get_flag_class(flag.type).compare(flag, submission):
            return True, 'Correct'
    return False, 'Incorrect'
You can’t perform that action at this time.