Simple encrypted credential management with GPG.
I have a lot of different sensitive environment variables to juggle. API keys, tokens, usernames, passwords, etc. I had been using simple shell scripts to set environment variables when needed, eg:
$ cat ~/Dropbox/creds/aws-work.sh
export AWS_ACCESS_KEY_ID=foo
export AWS_SECRET_ACCESS_KEY=bar
$ source ~/Dropbox/creds/aws-work.sh
$ echo $AWS_ACCESS_KEY_ID
foo
$ s3cmd ...
But I don't like storing these in plaintext on Dropbox.
Thus, how about a simple way to encrypt/decrypt these as needed with GPG?
Tested on Mac OSX 10.11 with gpg2
installed from homebrew, but should work
on most platforms with the above requirements.
Several options for installation, in order of recommendation:
$ brew install joemiller/taps/creds
$ brew install joemiller/taps/creds --HEAD
make install
:$ git clone https://github.com/joemiller/creds.git
$ cd creds ; make install
$ curl https://raw.githubusercontent.com/joemiller/creds/master/creds >./creds
$ chmod +x ./creds
If you're on OSX you may need to install GPG and create a keypair. You have a few options:
brew install gpg2 gpg-agent
Run gpg2 --gen-key
to generate a new keypair if you don't already have one.
brew uninstall creds
make install
: Run make uninstall
$ creds -h
usage: creds [-h|--help] [-v|--version] <subcommand> [arguments]
Simple encrypted credential file management with GPG.
The most commonly used subcommands are:
list list available credential stores
edit edit a credential store
import import an existing file into a new credential store
set display commands to set credentials from a credential store
unset display commands to unset credentials from a credential store
creds
reads configuration from ~/.credsrc
file, eg:
CREDS_DIR="$HOME/Dropbox/creds"
GPG_KEY=joeym@joeym.net
Required variables:
CREDS_DIR
: A directory where encrypted credentials files will be stored.GPG_KEY
: The GPG key to use for encrypting credentials. Use gpg -K
to
list keys.Optional variables:
GPG_BIN
: Path to GPG bin to use. If not set, creds
will look for gpg2
and gpg
in the path, preferring gpg2
if found.The edit
command will create a new credential store if one does not exist yet.
The format of credential stores is single line KEY=val
environment variable
style lines. All other lines will be ignored when using the set
and unset
commands.
$ creds edit aws-work
< .. $EDITOR launches .. >
AWS_ACCESS_KEY_ID=foo
AWS_SECRET_ACCESS_KEY=bar
$ creds list
Credential storage dir: /Users/joe/Dropbox/creds
- aws-work
- misc
- digitalocean
Use the set
command to print the contents of a credential store.
Usually you will wrap this with eval
to set the credentials in your shell's
environment.
$ creds set aws-work
export AWS_ACCESS_KEY_ID=foo
export AWS_SECRET_ACCESS_KEY=bar
You can then copy and paste to set these vars in your current shell or do it in one comand:
$ eval "$(creds set aws-work)"
$ echo $AWS_ACCESS_KEY_ID
foo
The export VAR
commands generated by creds set
will be prefixed with a single
whitespace character. If you're using zsh or bash as your shell and HISTCONTROL=
env var contains ignorespace
this will prevent the export statements (and your
secrets) from being stored in the command history. This is the default setting
in bash and zsh so it's probably already set correctly.
Use the run
command to load environment vars from a credential store and
execute a command with that environment.
Environment vars are added to the current environment so existing vars will also be available to the command.
$ creds run aws-work env | grep AWS
AWS_ACCESS_KEY_ID=foo
AWS_SECRET_ACCESS_KEY=bar
$ creds run aws-work s3cmd ls s3://some-bucket
Similar to, and inspired by, the excellent envchain util.
Use the unset
command to unset the credentials. This should also be used
with eval
.
$ creds unset aws-work
unset AWS_ACCESS_KEY_ID
$ eval $(creds unset aws-work)
$ cat ./circleci.keys
CIRCLE_TOKEN=foo
$ creds import ./circleci.keys
Encrypting './circleci.keys' to '/Users/joe/Dropbox/creds/circleci.keys.gpg'
creds is a great companion to direnv.
Place one or more eval statements in your .envrc
file:
$ mkdir some-aws-project
$ echo 'eval "$(creds set aws-personal)"' >some-aws-project/.envrc
$ cd some-aws-project
direnv: loading .envrc
direnv: export +AMAZON_ACCESS_KEY_ID +AMAZON_SECRET_KEY_ID
$ echo $AMAZON_ACCESS_KEY_ID
ABCDEFGHIJKLMNOP
$ cd ..
direnv: unloading
$ echo $AMAZON_ACCESS_KEY_ID
$
direnv is able to follow all of the env vars set by creds so when you leave the directory they will be automatically unloaded.
If you see an error such as the following (ambiguous) error during creds edit
:
gpg: [stdin]: encryption failed: Unusable public key
try using gpg --edit <key_id>
(or gpg2 --edit-key <key_id
for GPG 2.1+)
and marking the key as "ultimately trusted".
You may use variables in assignment for other variables in a cred set but you will need to enclose the eval statement with double quotes, example:
# creds edit foo
AWS_ACCESS_KEY_ID=ABCDEFGHIJKLMNOP
AMAZON_KEY_ID=$AWS_ACCESS_KEY_ID
$ eval $(creds set foo)
$ env | grep KEY_ID
AWS_ACCESS_KEY_ID=ABCDEFGHIJKLMNOP
AMAZON_KEY_ID=
$ eval "$(creds set foo)"
$ env | grep KEY_ID
AWS_ACCESS_KEY_ID=ABCDEFGHIJKLMNOP
AMAZON_KEY_ID=ABCDEFGHIJKLMNOP
Requirements:
bats
- (brew install bats
)shellcheck
- (brew install shellcheck
)Run make help
to get a list of tasks.
keybase
commands too? but don't introduce a
dependency on keybase.joe miller, 2016