Layouts in one call

batch reads a JSON array from stdin or --file and applies every entry.

windowctl batch --file work-mode.json --json
cat work-mode.json | windowctl batch

[
  { "app": "Google Chrome", "monitor": 2, "zone": "2B" },
  { "app": "Code",          "monitor": 1, "zone": "1A" },
  { "title": "Terminal",    "monitor": 1, "zone": "1B" },
  { "app": "Slack",         "monitor": 2, "zone": "2C" }
]

Each entry takes the same targeting as move: app or title to select, monitor plus zone, or explicit x/y/w/h.

Partial success is reported, not hidden

Every entry reports its own outcome. A layout where three windows moved and one was clamped tells you which one and why, and the exit code reflects it. That is the difference between a layout tool you can put in a login script and one you have to watch.

Why this instead of a loop

A loop of move calls re-enumerates every window on every iteration and gives you N separate exit codes to reason about. batch enumerates once and hands back one structured result — which is also what makes it the right primitive for an agent to call.