Decker : Declarative Penetration Testing Orchestration Framework

Decker is a penetration testing orchestration framework. It leverages HashiCorp Configuration Language 2 (the same config language as Terraform) to allow declarative penetration testing as code, so your tests can be versioned, shared, reused, and collaborated on with your team or the community.

Example of a decker config file:

// variables are pulled from environment
// they will be available throughout the config files as var.*
// ex: ${var.target_host}
variable “target_host” {
type = “string”
// resources refer to plugins
// resources need unique names so plugins can be used more than once
// they are declared with the form: ‘resource “plugin_name” “unique_name” {}’
// their outputs will be available to others using the form unique_name.*
// ex: nmap.443
resource “nmap” “nmap” {
host = “${var.target_host}”
plugin_enabled = “true”
resource “sslscan” “sslscan” {
host = “${var.target_host}”
plugin_enabled = “${nmap.443 == “open”}”

Run a plugin for each item in a list:

variable “target_host” {
type = “string”
resource “nslookup” “nslookup” {
dns_server = “”
host = “${var.target_host}”
resource “metasploit” “metasploit” {
for_each = “${nslookup.ip_address}”
exploit = “auxiliary/scanner/portscan/tcp”
options = {
RHOSTS = “${each.key}/32”
INTERFACE = “eth0”

Complex configuration combining for_each with nested values:

variable “target_host” {
type = “string”
resource “nslookup” “nslookup” {
dns_server = “”
host = “${var.target_host}”
resource “nmap” “nmap” {
for_each = “${nslookup.ip_address}”
host = “${each.key}”
// for each IP, check if nmap found port 25 open.
// if yes, run metasploit’s smtp_enum scanner
resource “metasploit” “metasploit” {
for_each = “${nslookup.ip_address}”
exploit = “auxiliary/scanner/smtp/smtp_enum”
options = {
RHOSTS = “${each.key}”
plugin_enabled = “${nmap[“${each.key}”].25 == “open”}”

Also Read – UserLAnd : Run a Linux Distribution or Application on Android

Output formats

Several output formats are available and more than one can be selected at the same time.

Setting DECKER_OUTPUTS_JSON or DECKER_OUTPUTS_XML to "true" will output json and xml formatted files respectively.

  1. Output .json files in addition to plain text: export DECKER_OUTPUTS_JSON="true"
  2. Output .xml files in addition to plain text: export DECKER_OUTPUTS_XML="true"

Running an example config with docker

Two volumes are mounted:

  • Directory named decker-reports where decker will output a file for each plugin executed. The file’s name will be {unique_resource_name}.report.txt.
  • examples directory containing decker config files. Mounting this volume allows you to write configs locally using your favorite editor and still run them within the container.

One environment variable is passed in:


This is referenced in the config files as {var.target_host}. Decker will loop through all environment variables named DECKER_*, stripping away the prefix and setting the rest to lowercase.

docker run -it –rm \
-v “$(pwd)/decker-reports/”:/tmp/reports/ \
-v “$(pwd)/examples/”:/decker-config/ \
-e \
stevenaldinger/decker:kali decker ./decker-config/example.hcl

When decker finishes running the config, look in ./decker-reports for the outputs.

Running an example config without docker

You’ll likely want to set the directory decker writes reports to with the DECKER_REPORTS_DIR environment variable.

Something like this would be appropriate. Just make sure whatever you set it to is an existing directory.

export DECKER_REPORTS_DIR=”$HOME/decker-reports”

You’ll also need to set a target host if you’re running one of the example config files.


Then just run a config file. Change to the root directory of this repo and run:

./decker ./examples/example.hcl


Using docker for development is recommended for a smooth experience. This ensures all dependencies will be installed and ready to go.

Refer to Directory Structure below for an overview of the go code.

Quick Start

  • (on host machine): make docker_build
  • (on host machine): make docker_run (will start docker container and open an interactive bash session)
  • (inside container): dep ensure -v
  • (inside container): make build_all
  • (inside container): make run

Initialize git hooks

Run make init to add a pre-commit script that will run linting and tests on each commit.

Plugin Development

Decker itself is just a framework that reads config files, determines dependencies in the config files, and runs plugins in an order that ensures plugins with dependencies on other plugins (output of one plugin being an input for another) run after the ones they depend on.

The real power of decker comes from plugins. Developing a plugin can be as simple or as complex as you want it to be, as long as the end result is a .so file containing the compiled plugin code and a .hcl file in the same directory declaring the inputs the plugin is expecting a user to configure.

The recommended way to get started with decker plugin development is by cloning the decker-plugin repository and following the steps in its documentation. It should only take you a few minutes to get a “Hello World” decker plugin running.

Installing plugins

By default, plugins are expected to be in a directory relative to wherever the decker binary is, at /internal/app/decker/plugins//.so. Additional paths can be added by setting the DECKER_PLUGIN_DIRS environment variable. The default plugin path will still be used if DECKER_PLUGIN_DIRS is set.

Example: export DECKER_PLUGIN_DIRS=”/path/to/my/plugins:/additional/path/to/plugins”

There should be an HCL file next to the .so file at /internal/app/decker/plugins//.hcl that defines its inputs and outputs. Currently, only string, list, and map inputs are supported. Each input should have an input block that looks like this:

input “my_input” {
type = “string”
default = “some default value”

Directory Structure

├── build
│ ├── ci/
│ └── package/
├── cmd
│ ├── decker
│ │ └── main.go
│ └──
├── deployments/
├── docs/
├── examples
│ └── example.hcl
├── githooks
│ ├── pre-commit
├── Gopkg.toml
├── internal
│ ├── app
│ │ └── decker
│ │ └── plugins
│ │ ├── a2sv
│ │ │ ├── a2sv.hcl
│ │ │ ├── main.go
│ │ │ └──
│ │ └── …
│ │ ├── main.go
│ │ ├──
│ │ └── xxx.hcl
│ ├── pkg
│ │ ├── dependencies/
│ │ ├── gocty/
│ │ ├── hcl/
│ │ ├── paths/
│ │ ├── plugins/
│ │ └── reports/
│ └──
├── Makefile
└── scripts

  • cmd/decker/main.go is the driver. Its job is to parse a given config file, load the appropriate plugins based on the file’s resource blocks, and run the plugins with the specified inputs.
  • examples has a couple example configurations to get you started with decker. If you use the kali docker image (stevenaldinger/decker:kali), all dependencies should be installed for all config files and things should run smoothly.
  • internal/pkg is where most of the actual code is. It contains all the packages imported by main.go.
    • dependencies is responsible for building the plugin dependency graph and returning a topologically sorted array that ensures plugins are run in a working order.
    • gocty offers helpers for encoding and decoding go-cty values which are used to handle dynamic input types.
    • hcl is responsible for parsing HCL files, including creating evaluation contexts that let blocks properly decode when they depend on other plugin blocks.
    • paths is responsible for returning file paths for the decker binary, config files, plugin config files, and generated reports.
    • plugins is responsible for determining if plugins are enabled and running them.
    • reports is responsible for writing reports to the file system.
  • internal/app/decker/plugins are modular pieces of code written as Golang plugins, implementing a simple interface that allows them to be loaded and called at run-time with inputs and outputs specified in the plugin’s config file (also in HCL). An example can be found at internal/app/decker/plugins/nslookup/nslookup.hcl.
  • decker config files offer a declarative way to write penetration tests. The manifests are written in HashiCorp Configuration Language 2) and describe the set of plugins to be used in the test as well as their inputs.