Shell Integration

The cd Problem

A compiled binary can't change the parent shell's working directory — it runs in a child process. wt solves this with a lightweight protocol between the binary and a shell wrapper function.

How It Works

1. All user-visible output (prompts, status messages, errors) goes to stderr. This keeps stdout clean for the protocol.
2. When a directory change is needed, the binary prints the target path to stdout and exits 0.
3. The shell wrapper captures stdout and cds into it if non-empty.

This means @clack/prompts interactive UI works normally (it writes to stderr/TTY), and the wrapper reliably gets the cd target from stdout.

Shell Wrapper

Add this to your .zshrc or .bashrc:

wt() {
  local cd_target
  cd_target=$(command wt "$@")
  local exit_code=$?
  [[ $exit_code -ne 0 ]] && return $exit_code
  [[ -n "$cd_target" ]] && cd "$cd_target"
}

How the wrapper works

• command wt "$@" runs the actual binary (bypassing the function itself).
• cd_target=$(...) captures stdout into a variable.
• stderr passes through to your terminal normally — you see all prompts and log messages.
• If the binary exits non-zero, the wrapper returns the same exit code.
• If stdout is non-empty, the wrapper cd's into the path.

Which Commands Produce cd Targets

Command	Outputs to stdout?	When?
wt add	Yes	Path to new worktree (unless --no-cd)
wt checkout / switch	Yes	Path to new worktree (unless --no-cd)
wt cd	Yes	Path to worktree or repo root
wt rm	Yes	Repo root (only if cwd was inside removed worktree)
wt ls	No	Output goes to stderr
wt purge	No	Output goes to stderr
wt completion	Yes	Completion script (not a cd target — used with source)

Fish Shell

Fish uses a slightly different syntax:

function wt
  set -l cd_target (command wt $argv)
  set -l exit_code $status
  if test $exit_code -ne 0
    return $exit_code
  end
  if test -n "$cd_target"
    cd $cd_target
  end
end