Skip to content

Latest commit

 

History

History
667 lines (516 loc) · 48.2 KB

README.md

File metadata and controls

667 lines (516 loc) · 48.2 KB

Dotfiles

Brendon Smith (br3ndonland)

Table of Contents

Overview

What

This repo contains dotfiles, which are application configuration and settings files. They frequently begin with a dot, hence the name. Dotfiles are compatible with Linux and macOS.

Why

  • Make developer environments automated and disposable. Disposability is an important concept in infrastructure-as-code DevOps, serverless computing, CI/CD, and more recently, in-browser development environments. Why aren't developers applying automation and disposability to their own computers? With an automated disposable developer environment, setup of a new machine is fast and easy. This approach is also liberating - I can purchase a new computer (or wipe an existing one), run bootstrap.sh, and be up and running again in no time.
  • Know when and why settings change. I not only know what tools and settings I'm using, but when and why I chose the tools and settings. This has been particularly important for VSCode, because settings change (and break) frequently, and it helps to record troubleshooting info in the Git log.
  • Learn new skills. I learn skills, like shell scripting, that are useful and don't go out of date quickly. I wouldn't know shell as well if I didn't work on my developer environment. I learn these skills by tinkering a little bit at a time, in an unstructured way. It's time I might not otherwise be writing code.

How

This dotfiles repository is meant to be installed by bootstrap.sh.

bootstrap.sh is a shell script to automate setup of a new macOS or Linux development machine. It is idempotent, meaning it can be run repeatedly on the same system. To set up a new machine, simply open a terminal and run the following command:

STRAP_GIT_EMAIL="you@example.com" STRAP_GIT_NAME="Your Name" STRAP_GITHUB_USER="username" \
  /usr/bin/env bash -c "$(curl -fsSL https://raw.githubusercontent.com/br3ndonland/dotfiles/HEAD/bootstrap.sh)"

The following environment variables can be used to configure bootstrap.sh, and should be either set before with export, or inline within the command to run the script:

  • STRAP_GIT_EMAIL: email address to use for Git configuration. Will error and exit if not set.
  • STRAP_GIT_NAME: name to use for Git configuration. Will error and exit if not set.
  • STRAP_GITHUB_USER: username on GitHub or other remote from which dotfiles repo will be cloned. Defaults to my GitHub username, so you should set this if you're not me.
  • STRAP_DOTFILES_URL: URL from which the dotfiles repo will be cloned. Defaults to https://github.com/$STRAP_GITHUB_USER/dotfiles, but any Git-compatible URL can be used, so long as it is accessible at the time the script runs.
  • STRAP_DOTFILES_BRANCH: Git branch to check out after cloning dotfiles repo. Defaults to main.

There are some additional variables for advanced usage. Consult the bootstrap.sh script to see all supported variables.

bootstrap.sh will set up macOS and Homebrew, run scripts in the scripts/ directory, and install Homebrew packages and casks from the Brewfile. A Brewfile is a list of Homebrew packages and casks (applications) that can be installed in a batch by Homebrew Bundle. The Brewfile can even be used to install Mac App Store apps with the mas CLI. Note that you must sign in to the App Store ahead of time for mas to work.

The following list is a brief summary of permissions related to bootstrap.sh.

  • Initial setup of Homebrew itself does not require an admin user account, but does require sudo. See the Homebrew installation docs, Homebrew/install#312, and Homebrew/install#315.
  • After Homebrew setup, use of sudo with brew commands is discouraged.
  • After Homebrew setup, commands such as brew bundle install --global should be run from the same user account used for setup. Attempts to run brew commands from another user account will result in errors, because directories that need to be updated are owned by the setup account. If access to the setup account is not routinely available, an alternative approach could be to change ownership of Homebrew directories to a group that includes the user account used for Homebrew setup as well as other users that need to run Homebrew commands.
  • bootstrap.sh can run with limited functionality on non-admin and non-sudo user accounts. A plausible use case could exist in which an admin runs bootstrap.sh to configure the system initially, then a non-admin runs bootstrap.sh to configure their own account. In this use case, the non-admin user should not need admin or sudo privileges, because all the pertinent setup (FileVault disk encryption, XCode developer tools, Homebrew, etc) is already complete.

Users with more complex needs for multi-environment dotfiles management might consider a tool like chezmoi.

Hardware

macOS

  • macOS setup is automated with macos.sh.

  • Karabiner Elements is used for keymapping.

    • Settings are stored in .config/karabiner/karabiner.json. Note that Karabiner will auto-format the JSON with four spaces. To avoid changing the formatting with the Prettier autoformatter, I added karabiner.json to .prettierignore.

    • Simple modifications:

      From key To key
      caps_lock escape
      escape caps_lock
    • Complex modifications:

      • Launch Terminal with Cmd+Escape
      • See karabiner.json for more.
    • Devices

      • Disable built-in keyboard when external keyboard is connected

Homebrew package management

  • Homebrew is a package manager that includes Homebrew-Cask to manage other macOS applications. See the Homebrew docs for further info.

  • The list of "formulae" (packages), "casks" (apps), and mas apps (Mac App Store apps) is stored in Brewfile. The Brewfile works with Homebrew Bundle to manage all Homebrew packages and casks together.

  • Key Brew Bundle commands:

    # Install or update everything in the Brewfile
    brew bundle install --global
    # Check for programs listed in Brewfile
    brew bundle check --global
    # Remove any Homebrew packages and casks not in Brewfile
    brew bundle cleanup --force --global

Git

  • Git is the version control system used on GitHub. Why use Git? Git enables creation of multiple versions of a code repository called branches, with the ability to track and undo changes in detail. If you're new to Git, the Git Book is helpful.
  • I install Git with Homebrew.
  • I configure Git to connect to GitHub with SSH.
  • I store Git configuration in .gitconfig.
  • I sign Git commits and tags with 1Password and SSH. In the past, I have also signed with GPG.
  • Why sign Git commits and tags?
    • Signing verifies user identity. In case you haven't heard, anyone can use Git to impersonate you. Signing helps avoid impersonation attacks. The simplest form of signing can be seen when performing Git operations through the GitHub UI with valid credentials (making a commit, merging a PR, etc). As the GitHub docs explain, these operations are signed with GitHub's key (signature 4AEE18F83AFDEB23). Operations signed with this key indicate valid GitHub credentials, so signing with GitHub's key is somewhat like one-factor authentication. To make this more secure, users can use their own keys, which involves generating and maintaining custody of a key, setting up Git to sign commits and tags with the key, and adding the public key to the GitHub account. This can be even more secure, somewhat like two-factor authentication, because it indicates that the user has both valid GitHub credentials and the valid private key used for signing. Signing with a user-generated key also allows attestation. If you navigate to my GitHub profile, you can see several SSH and PGP key signatures that I have associated with my identity on GitHub. This means, "I attest that I possess the corresponding private keys and use them for signing." Commits and tags signed with those keys show up as "verified" on GitHub.
    • Signing verifies changes. This is especially important for tags because tags are often used to perform releases. Signed tags indicate that the release comes from a trusted source, because the user possessing the signing key has verified the contents of the release.
  • See the GitHub docs for further info on signing.

Shell

Text editors

VSCode desktop

I write code with VSCodium, an alternate build of Microsoft Visual Studio Code (VSCode) that is free of proprietary features and telemetry.

VSCode settings, keybindings, and extension lists are stored in this repo. Extensions can be installed by running vscode-extensions.sh along with the editor command name, like vscode-extensions.sh codium for VSCodium. The script uses the VSCode extension CLI.

VSCode offers other options for managing settings. The Shan.code-settings-sync extension was popular in the past, and stored settings in a GitHub Gist. VSCode now offers a built-in settings sync feature (introduced in VSCode 1.48 July 2020). This repo uses Git for settings sync instead of VSCode's settings sync feature. Why not use VSCode settings sync?

  • VSCode settings sync requires a Microsoft or GitHub login. What if you use GitLab?
  • Where does VSCode settings sync actually store the data? The Shan.code-settings-sync extension stored data in a GitHub Gist, so it was possible to view the synced settings directly. VSCode only allows the synced settings to be viewed from within VSCode. There's no Git repo or Gist where you can go to see line-by-line changes.
  • VSCode settings sync uses a "Merge or Replace" dialog box and a pseudo-Git merge conflict resolver. It can be complicated and confusing to use.

VSCode profiles (introduced in VSCode 1.76 February 2023) assist with switching among different editor environments, like personal and work computers. This repo does not use VSCode profiles, but uses a simple Git branching strategy for managing these environments. Why not use VSCode profiles?

  • Most settings are the same across all profiles, but VSCode profiles store separate JSON settings for every profile
  • Exporting a profile to a file exports the settings as an inline dumped JSON string, making version control and reading difficult
  • Exporting a profile to a file includes metadata like snippets.usageTimestamps, which is apparently tracking every time snippets are used. Why? Why would you need to track this in a profile? This also makes version control more difficult because it creates a Git diff every time a snippet is used.
  • The VSCode profiles docs describe the use case as, "Since profiles are remembered per workspace, they are a great way to customize VS Code for a specific programming language... Using this approach, you can easily switch between workspaces and always have VS Code configured the right way." There are already workspace and repo settings available that do the same thing.

VSCode browser

codespaces

I have tried GitHub's cloud-hosted VSCode, called Codespaces. So far, I don't like it because:

  • Keybindings conflict with browser keybindings. The command palette may not be able to be opened with the usual keybindings, because Cmd+shift+p is already in use by the browser. This was partially addressed by the VSCode extension, which allows remote development from outside the browser.
  • Lag. There's lag between when a character is typed and when it appears. This can be very bothersome.
  • Files. It's unclear if or how persistent file storage is available. Local filesystem access is limited.
  • GPG commit signing has some limitations. GitHub offers commit signing with GPG, but private keys must be in GitHub's custody.
  • Customization is complicated.
    • Codespaces doesn't automatically clone or run these dotfiles. They claim that codespaces will automatically clone your dotfiles and run bootstrap.sh, but the repo is not cloned to ~/.dotfiles, and bootstrap.sh is not running, even after enabling the "Automatically install dotfiles" setting in user preferences.
    • Codespaces can't read VSCode settings from dotfiles. The docs explain, "Currently, Codespaces does not support personalizing the User settings for the Visual Studio Code editor with your dotfiles repository." The separate Settings Sync service is required to sync settings.
    • Extensions have to be managed separately from VSCode desktop. There is a "github.codespaces.defaultExtensions" setting that requires its own list of extensions, and it's unclear if Codespaces can install extensions from VSIX (like Dracula Pro).
    • Codespaces may not be able to install and use custom fonts (like Dank Mono). Font capabilities may be limited by the browser.
    • For further customization, Codespaces supports "dev containers," which are created with special Dockerfiles in each project. This adds further complication and maintenance overhead.
  • Codespaces pricing requires enterprises to set up separate spending limits, and is not specified for public repos on free plans. Do you want to pay for your text editor by the minute?

code-server

  • code-server is an open-source alternative to codespaces.

  • I set up code-server on a Linux cloud server. I prefer to use DigitalOcean, following their recommended initial setup guide for "droplets" (VMs):

    • Set up SSH agent forwarding on local machine to avoid having to deposit SSH private keys on droplet
    • Add SSH public key when creating droplet
    • Add a user data script like linux-userdata.sh
  • Installation and setup on the server:

    curl -fsSL https://code-server.dev/install.sh | sh
    sudo systemctl enable --now code-server@$USER
  • Server configuration file: note that user-data-dir must be an absolute path if running with the systemctl background service.

    bind-addr: 127.0.0.1:8080
    auth: none
    password: false
    cert: false
    user-data-dir: /home/brendon/.dotfiles/vscode
  • Local machine

    • Forward port from server by running ssh -N -L 8080:127.0.0.1:8080 code-server on the local machine.
    • Open http://localhost:8080 in a browser and you should see VSCode!
    • Multiple workspaces can be opened by passing the ?workspace= query parameter in the URL. Each browser tab can have a workspace open.
    • Proxy ports back to local machine with the /proxy endpoint. For example, to hit an API endpoint running on port 1025 on the server, curl :8080/proxy/1025. As explained in the VSCode docs, if developing on Linux, note that non-root users may not be able to expose ports less than 1024. The port is set to 1025 in the debugger config for this reason.
  • Extensions:

    • code-server has its own extension marketplace that is created by scraping GitHub.
    • You can also install extensions with the CLI: code-server --install-extension.
  • Settings

    • If the shell doesn't look right: Command Palette -> Terminal: Select Default Shell
    • The browser may grab some VSCode keybindings. I prefer to use Safari, because it grabs the least shortcuts.
    • code-server can't be used as a Git editor, as far as I can tell. It can open text files from the command-line, but the --wait switch is not recognized.
    • The clipboard doesn't completely work. See coder/code-server#1106.
    • The font can't yet be customized directly. See coder/code-server#1374.
    • You can change the color theme, but you may need to re-select the theme each time you open a browser tab.

Fonts

I use Recursive Mono. It's available for download from Homebrew (brew install --cask font-recursive-code) or GitHub. Fira Code and Ubuntu Mono are decent free alternatives.

I previously used Dank Mono. The rounded characters are eminently readable, the italics are elegant, and the ligatures are intuitive. Unfortunately, it's not correctly monospaced. I reached out to the Dank Mono creator Phil Pluckthun after it stopped working in kitty 0.36:

Hi Phil,

I've been using Dank Mono for years and love it, but I'm having an issue in which the font is not detected as monospaced on macOS. I can still use Dank Mono in VSCode, but some other terminal applications and text editors only accept monospaced fonts. I've tried modifying the macOS Font Book "Fixed Width" smart collection so it includes Dank Mono, but it's still not picked up as monospaced. Is there something I can do about this?

It may have to do with the ligatures. If I load Dank Mono into a font editor, I can see that the ligatures are not the same width as the other characters. See the attached screenshot.

It may also have to do with spacing. The italic variant lacks a spacing property.

~
❯ fc-list : family style spacing outline scalable | grep Dank
Dank Mono:style=Bold:spacing=100:outline=True:scalable=True
Dank Mono:style=Italic:outline=True:scalable=True
Dank Mono:style=Regular:spacing=100:outline=True:scalable=True

Related:

Thanks for your help,

Brendon Smith

Phil's reply:

Hiya,

Some terminal applications require a specific monospace flag to be set in the font file that was only gaining traction after the font was released. Some also just don't support fonts with ligatures. In Dank Mono's case the specific flag that some applications require to select the font as a monospaced one simply isn't set, and a different, older one is set instead.

Cheers & Best,

Phil

Language-specific setup

JavaScript

  • node is a JavaScript runtime used to run JavaScript outside of a web browser.
  • npm is a package manager written in node.js, included when node is installed.
    • It's difficult to keep track of global npm packages. There's no easy way to do it with the usual package.json. As Isaac Schlueter commented in 2012,

      Yeah, we're never going to do this.

    • Instead, packages can be installed with Homebrew, or with npm-globals.sh.
    • npm-check can be used to manage global packages after install, with npm-check -ug. If not using npm-check, a list of global npm packages can be seen after installation with npm list -g --depth=0.
  • I use the Prettier autoformatter and the Prettier VSCode extension to format my web code, including JavaScript and Vue.js. Prettier is an extremely helpful productivity tool, and I highly recommend it. Autoformatters save time and prevent bikeshedding.
  • Compared with Prettier, ESLint formats less code languages, requires complicated setup, and doesn't work well when installed globally.
  • In the past, I also used JavaScript Standard Style (aka StandardJS). Standard Style has also reportedly been favored by Brendan Eich (creator of JavaScript) and Sir Tim Berners-Lee (creator of the World Wide Web). Prettier provides a similar code style, but with more features, so I use Prettier instead.

Python

PGP

  • I use Gnu Privacy Guard (GPG, the free implementation of Pretty Good Privacy (PGP)), Keybase, and Proton Mail to encrypt and share messages, passwords, and other sensitive info.
  • PGP vs SSL: SSL/TLS/HTTPS encrypts data in transit, but the storage provider like Dropbox, Google, or Slack can still read it. Communications which are end-to-end PGP encrypted can only be read by the sender or recipient, never the provider.

GPG

GPG is an implementation of OpenPGP.

Installation

  • Install gnupg:
    • macOS: brew install gnupg
    • Ubuntu Linux: sudo apt-get install gnupg
    • Manually: Download from GnuPG
  • Install pinentry:
    • macOS: brew install pinentry (terminal) or brew install pinentry-mac (app from GPGTools suite that enables use of macOS keychain and GUI apps like VSCode)
    • Ubuntu Linux: apt install pinentry
    • Manually: On macOS, Apple's command-line tools (xcode-select) should have the build prerequisites. Download libgpg-error, libassuan, and pinentry, then build from source and install in order (libgpg-error, then libassuan, then pinentry):
      cd /path/to/libgpg-error
      ./configure; make; sudo make install
      cd /path/to/libassuan
      ./configure; make; sudo make install
      cd /path/to/pinentry
      ./configure; make; sudo make install
  • Set the pinentry program:
    • ~/.gnupg/gpg-agent.conf: pinentry-program /opt/homebrew/bin/pinentry-tty
    • The path apparently has to be absolute, so it may vary by system.
    • See the GPG agent options docs for more.
    • pinentry-tty allows plain-text password entry. pinentry may raise confusing "Screen or window too small" errors in some terminals.
  • Export the GPG_TTY environment variable: export GPG_TTY=$(tty) (can add to shell profile to automatically export). Note that this is not the same thing as export GPG_TTY=$TTY, which may raise cryptic Inappropriate ioctl for device errors.
  • Ensure proper permissions are set on GPG config files:
    chmod 700 ~/.gnupg
    chmod 600 ~/.gnupg/gpg.conf
  • See the GPG configuration docs for more.

GPG and YubiKey

Resources:

A YubiKey has admin and non-admin OpenPGP PINs.

  • Default OpenPGP admin PIN is 12345678.
  • Default OpenPGP non-admin PIN is 123456.
  • Change OpenPGP PINs with gpg --change-pin or ykman openpgp access change-admin-pin/ykman openpgp access change-pin.

To use a GPG key on a YubiKey with a new computer, plug in the YubiKey, check the status, and fetch the public keys.

~
❯ gpg --card-status

~
❯ gpg --card-edit

gpg/card> admin
Admin commands are allowed

gpg/card> url
URL to retrieve public key: https://github.com/<YOUR_GITHUB_USERNAME>.gpg

gpg/card> fetch

gpg/card> quit

To move a GPG key from a computer to a YubiKey, use gpg --edit-key followed by keytocard. The keytocard command is repeated for each "where to store the key" option. After the two keytocard commands, there may be confusing output regarding whether or not to "save" your changes, including Note: the local copy of the secret key will only be deleted with "save". As recommended in Yubico support: using your YubiKey with OpenPGP, entering "no" at these prompts will avoid deleting the GPG key from the computer (but will still save it to the YubiKey).

~
❯ gpg --edit-key <KEY_ID>

gpg> keytocard
Really move the primary key? (y/N) y
Please select where to store the key:
   (1) Signature key
   (3) Authentication key
Your selection? 1
gpg: pinentry launched (11284 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter your passphrase, so that the secret key can be unlocked for this session
Passphrase:
gpg: pinentry launched (11285 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter the Admin PIN

Number: 12 345 678
Holder:
Admin PIN:
gpg: pinentry launched (11287 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter the Admin PIN

Number: 12 345 678
Holder:
Admin PIN:

sec  ed25519/16digit_PGPkeyid
     created: 1900-01-01  expires: never       usage: SC
     trust: unknown       validity: unknown
ssb  cv25519/16digit_PGPkeyid
     created: 1900-01-01  expires: never       usage: E
[ unknown] (1). you <you@example.com>

Note: the local copy of the secret key will only be deleted with "save".

gpg> keytocard
Really move the primary key? (y/N) y
Please select where to store the key:
   (1) Signature key
   (3) Authentication key
Your selection? 3

gpg: pinentry launched (11284 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter your passphrase, so that the secret key can be unlocked for this session
Passphrase:
gpg: pinentry launched (11285 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter the Admin PIN

Number: 12 345 678
Holder:
Admin PIN:
gpg: pinentry launched (11287 tty 1.2.1 /dev/ttys004 xterm-kitty - 20620/501/4 501/20 0)
Please enter the Admin PIN

Number: 12 345 678
Holder:
Admin PIN:

sec  ed25519/16digit_PGPkeyid
     created: 1900-01-01  expires: never       usage: SC
     trust: unknown       validity: unknown
ssb  cv25519/16digit_PGPkeyid
     created: 1900-01-01  expires: never       usage: E
[ unknown] (1). you <you@example.com>

Note: the local copy of the secret key will only be deleted with "save".

gpg> Q
Save changes? (y/N) N
Quit without saving? (y/N) y

GPG key generation

  • Run gpg --full-generate-key from the command line to generate a key. Respond to the command-line prompts. The maximum key size of 4096 is recommended.
  • View keys with gpg --list-secret-keys.
  • Run gpg --send-keys <keynumber> to share the public key with the GPG database. It takes about 10 minutes for the key to show up in the GPG database.

Key import and export

  • Import a GPG key from a file: gpg --import /path/to/privatekey.asc
  • Export your GPG public key:
    • Copy to clipboard (for pasting into GitHub/GitLab): gpg --armor --export | pbcopy
    • Export to a file: gpg --armor --export > public.gpg

Sending messages

  • Locate another user's key in the global database with gpg --search-keys <email>.
  • Encrypting communications
    • Encrypt a message with echo "Hello, World!" | gpg --encrypt --armor --recipient "<email>". Optionally, save the encrypted message in a .gpg file.
    • If the message was saved in a file, send the file over email, Slack, or any other medium.
    • Decrypt the message with gpg --decrypt.
      • If copying the encrypted text directly, include it in quotes: echo "BIG LONG GPG STRING" | gpg --decrypt.
      • If reading a file, include the filename when decrypting: gpg --decrypt message.gpg.
      • Decrypted output can be autosaved to a new file: gpg --decrypt message.gpg --output file.txt.

Signing Git commits with GPG

Note that SSH can also be used to sign Git commits. See the SSH section for further details.

  • See Pro Git: Signing your work.
  • Install and configure pinentry as described above.
  • Configure Git to use GPG and your key for commits, using .gitconfig:
    • Set signingkey: git config --global user.signingkey 16digit_PGPkeyid the 16 digit PGP key id is the partial 16 digit number listed on the sec line).
      [user]
      name = your name
      email = you@email.com
      signingkey = 16digit_PGPkeyid
    • Turn on gpgsign:
      [commit]
      gpgsign = true

General

  • Start the agent: gpg-connect-agent /bye
  • Reload the agent configuration: gpg-connect-agent reloadagent /bye
  • Stop the agent: gpgconf --kill gpg-agent or gpgconf --kill all. See the GPG docs on invoking gpg-agent.
  • Verify GPG signing capabilities: echo "test" | gpg --clearsign
  • Trust GPG keys using the GPG TTY interface:
    • If you see gpg: WARNING: This key is not certified with a trusted signature! when examining signed Git commits with git log --show-signature, you may want to trust the keys.
    • Enter the GPG key editor from the command line with gpg --edit-key <PGPkeyid>.
    • Set trust level for the key by typing trust, hitting enter, and entering a trust level.
    • See the GPG docs for more info.
  • GitHub GPG instructions
  • GitLab GPG instructions
  • If working on a server, you can use ssh agent forwarding to access your SSH and GPG keys without having to copy them.

Keybase

Zoom acquisition

Keybase was acquired by Zoom, and its future is uncertain.

Useful features

  • PGP key and identity management
    • Keybase solves the key identity problem. Even if you have someone's public PGP key, you can't verify it actually came from them unless you exchange it in person. Keybase provides a unified identity for verification of PGP keys. Each device gets its own private key, and they share identity. It was previously challenging to move PGP keys among devices, but now it can be accomplished simply by signing in to Keybase.
    • Following someone is a way of verifying their cryptographic identity, not a way of subscribing to updates from the person like social media.
    • PGP key features are available through the CLI with keybase pgp commands.
  • Chat
  • Teams
  • Encrypted files
    • The Keybase file system (KBFS) is like an encrypted Dropbox or Google Drive cloud storage system.
    • Integrates with the macOS finder through use of the FUSE for macOS package.
  • Crypto tools
  • Git
    • Keybase allows users and teams to create and store end-to-end encrypted Git repositories. See the Keybase Git docs and Keybase Git blog post.
    • Treat Keybase Git repos as remotes (like GitHub repos). They can be cloned, pushed, and pulled, as you would do for GitHub repos. For example: git remote add keybase keybase://$PUBLIC_OR_PRIVATE/$USERNAME/$REPONAME.
    • As of Keybase 5.1.0, Git LFS is also enabled.
    • Keybase can't be used to sign Git commits. The best method is to export your PGP key from Keybase to GPG, and then sign Git commits with GPG.
    • There has been some debate about the need to sign Git commits at all. Linus Torvalds has recommended the use of git tag -s to sign with tags instead. The Keybase developers sign releases with tags, but don't always sign commits to the Keybase source code. However, in order to sign tags, you still need to set up commit signing, so why not just sign commits also? Whether you sign all commits or just tags, Keybase should improve this feature.

Proton Mail

I use Proton Mail for PGP-encrypted email.

SSH

Key generation

To generate an SSH key:

ssh-keygen -t ed25519 -C "your_email@example.com" -f ~/.ssh/id_ed25519_"$(id -un)"

If you have a FIDO2 security key that supports discoverable credentials (formerly known as resident keys), such as a YubiKey, you can generate an SSH key that is stored directly on the FIDO2 hardware device.

You'll need libfido2 and OpenSSH version 8.2 or later.

brew install libfido2 openssh

Next, generate a key with the -O resident option. You can additionally set a PIN on the YubiKey, which requires the YubiKey Manager CLI or GUI, and require PIN verification for use of the SSH key, with the -O verify-required option. In this scenario, the SSH key itself does not need a password. The password is replaced by the YubiKey and its PIN.

ssh-keygen -t ed25519-sk -O resident -O verify-required -C "your_email@example.com" -f ~/.ssh/id_ed25519_"$(id -un)"

Connecting to GitHub

See the GitHub docs on connecting to GitHub with SSH.

GitHub supports use of SSH keys from FIDO2 security key hardware devices like YubiKeys. See the GitHub docs, GitHub blog, and Yubico blog.

GitHub also supports use of SSH keys for signing Git commits. See the GitHub changelog and GitHub docs. See the 1Password section for instructions.

SSH agent forwarding

If working on a server, you can use ssh agent forwarding to access your SSH and GPG keys without having to copy them.

Host yourserver.com
  ForwardAgent yes

1Password SSH features

1Password includes features for managing SSH keys. At this time, SSH features are limited to your Personal vault.

To get started:

  • Generate or import an SSH key
  • Upload the key to GitHub or any platform to which you connect with SSH
  • Turn on the 1Password SSH agent
  • Update the SSH config to use the 1Password IdentityAgent
  • Optionally, simplify the agent path by creating a symlink to ~/.1password/agent.sock.

1Password also supports Git commit signing with SSH keys. See the 1Password blog and GitHub changelog.

To sign Git commits with SSH and 1Password:

  • Tell GitHub about the SSH key:
    • Go to https://github.com/settings/keys
    • Click "New SSH key"
    • Select the key type "signing key"
    • Allow the 1Password browser extension to autofill the "key" input field with an SSH public key. Either generate a new SSH key with 1Password or use an existing one. The same SSH key can be used for both authentication and signing.
  • Tell Git about the SSH key:
    • Set git config gpg.format = ssh
    • Set git config gpg.ssh.allowedsignersfile=~/.ssh/allowed_signers
    • Set git config gpg.ssh.program=/Applications/1Password.app/Contents/MacOS/op-ssh-sign
    • Set git config user.signingkey to the SSH public key
    • Create the file ~/.ssh/allowed_signers (when using this repo, will be symlinked from ~/.dotfiles/.ssh/allowed_signers)
    • For each signing key, add a single line to the ~/.ssh/allowed_signers specifying the combination of git config user.email and git config user.signingkey, in that order

General productivity

Media

Science

(Back to top)