A job (as part of a workflow) runs inside a virtual machine. You choose one of the environments provided by them (e.g. ubuntu-latest or windows-2019).
A job consists of one or more steps. A step may be a simple shell command, using run. But it may also be an action, using uses
name: CI
on: [push]
jobs:
myjob:
runs-on: ubuntu-18.04 # linux required if you want to use docker
steps:
# Those steps are executed directly on the VM
- run: ls /
- run: echo $HOME
- name: Add a file
run: touch $HOME/stuff.txt
# Those steps are actions, which may run inside a container
- uses: actions/checkout@v1
- uses: ./.github/actions/my-action
- uses: docker://continuumio/anaconda3:2019.07
run: <COMMAND> executes the command with the shell of the OS
uses: actions/checkout@v1 runs the action from the user / organization actions in the repository checkout (https://github.com/actions/checkout), major release 1
uses: ./.github/actions/my-action runs the action which is defined in your own repository under this path
uses: docker://continuumio/anaconda3:2019.07 runs the anaconda3 image from user / organization continuumio, version 2019.07, from the Docker Hub (https://hub.docker.com/r/continuumio/anaconda3)
Keep in mind that you need to select a linux distribution as the environment if you want to use Docker.
Take a look at the documentation for uses and run for further details.
It should also be noted that there is a container option, allowing you to run any steps that would usually run on the host to be runned inside a container: https://help.github.com/en/articles/workflow-syntax-for-github-actions#jobsjob_idcontainer