dotfiles

A personal collection of dotfiles and scripts for syncing settings across machines.

The primary repository lives on tangled.org, which gets synced to github.com.

Setup

Install and symlink dotfiles into place:

just install

Just commands

This repo uses just as a task runner. Run just or just --list to see all recipes.

• just link — symlink dotfiles via link.sh
• just brew — install Homebrew packages from Brewfile
• just node — install/pin ambient Node via mise
• just install — brew, link, mise node, and upgrade toolchains
• just moshi-setup <token> — pair and start moshi-hook (one-time)

Tools

Node (mise)

Ambient node is provisioned by mise, not nvm/fnm or nub's shim. Global pin and settings live in mise/config.toml (linked to ~/.config/mise/config.toml). Interactive shells activate mise from .zshrc; login shells / IDEs get shims from .zprofile.

just node                 # install + pin global node@26
mise use node@22          # pin a project (writes mise.toml)
mise doctor               # verify activate/shims

Project pins also come from .node-version, .nvmrc, or package.json#devEngines when those are present. Nub stays the package manager (nub / nubx / nub install) and uses the mise-provided node on PATH.

Moshi

moshi-hook bridges local coding agents to the Moshi iOS app (inbox events, approvals, Live Activity). It's installed via the Brewfile; finish the one-time setup per machine with:

just moshi-setup <pairing-token> # token from Moshi app → Settings → Hooks

No tmux here — sessions run in herdr, which moshi-hook detects automatically via $HERDR_ENV. (Don't call moshi <dir>; that alias launches a tmux session.)

The moshi/ directory holds the Dark Greeny custom app theme and its import docs.

Crane

Useful tool for managing containers.

• Docs
• Common Commands

fx

fx is an experimental native coding agent CLI (~6.4 MB Zig binary, currently v0.0.4). State stays in ~/.fx — fx manages that folder itself; do not symlink it into a repo, fx deliberately refuses to write durable state into git-tracked paths.

just install-fx                    # install/upgrade to FX_VERSION (default v0.0.4)
FX_VERSION=v0.0.5 just install-fx  # pin a different release
fx                                 # start an interactive session in the cwd
fx ask "explain this error"        # one noninteractive request, then exit
fx doctor                          # post-install health check (auth, state, git)
fx status                          # resolved model / permission mode / workspace
fx upgrade                         # binary upgrade on selected release channel

Auth once per host (fx login for Vercel, or fx setup for an AI Gateway API key) and ~/.fx/settings.json persists model, permission mode, MCP servers, and prompt history.

Resume sessions, record transcripts, and add per-process directories with global flags:

fx --resume last                   # resume the last session in this workspace
fx -c                              # alias for --continue / --resume-last
fx --record                        # save terminal output under ~/.fx/
fx --add-dir ../shared             # grant the session access to another dir

Slash commands inside the interactive shell: /help, /settings, /permissions, /sandbox, /workspace, /status, /usage. Repo-safe config (sandbox hints, additional directories, permission overrides) goes in the workspace's own .fx.json, not ~/.fx.

Author

Created by frytg.digital

License

MIT