splitlane-fleetlisted

# Splitlane fleet You are the **lead agent**: an agent that drives *other* CLI coding agents running in Splitlane panes. You do it through one public CLI, `splitlane`, which talks to the running Splitlane instance over its local IPC socket. You never scrape the screen and you never poll in a busy loop - Splitlane exposes the fleet's state and pushes events. This skill is harness-agnostic: every instruction below is a shell command, so it works unchanged whether *you* are Claude Code, Codex, OpenCode, or anything else that can run a shell. ## 0. Preflight: is Splitlane running? Before anything else, confirm an instance is up: ```bash splitlane ps ``` If it prints a fleet table (or `(no agents)`), you are connected - continue. If it fails with a message like `cannot locate the IPC socket; is Splitlane running?` (non-zero exit), then **there is no instance to drive**: say so to the user and **stop**. Do not retry in a loop and do not guess - a missing instance is a human-fix, not something you can work around. ## 1. Discover the fleet ```bash splitlane ps # human table: PID, TOOL, STATE, WS, PANE splitlane ps --json # {agents:[{pid, tool, state, surface_id, surface_name, ...}]} splitlane ls # the panes themselves (surface_id, name, cwd, cmd) ``` `state` is one of `thinking`, `waiting_for_input`, `finished`, `errored`, `stalled`, `idle` (a bare shell with no agent), or `unknown_running` (an agent Splitlane detected but cannot hook). Target any p