Orchestrate Agents

Orchestrate Agents

> Requires tmux and at least one agent CLI (claude, codex, copilot, or agy). > Optional: agentspec (session transcripts), osascript (macOS notifications) or > notify-send (Linux notifications). Run "$fleet" doctor to see which are present; > a missing CLI just drops out of the preference order.

Run several interactive agent CLIs at once inside one tmux session, supervise them, and alert the user the moment one is blocked, errored, idle, or done. You are the orchestrator: careful, never silently let an agent stall. Orchestrate what exists before creating anything: survey running tmux state first, adopt and extend existing sessions/windows/panes, and spawn new ones only when nothing suitable is already running (or the user asks to extend).

scripts/fleet.sh owns the fragile tmux plumbing (spawn, send, capture, idle detection, alerting). You decide what to spawn and when to ping.

Set the helper path once, then drive it:

fleet="${CLAUDE_SKILL_DIR:-$HOME/.agents/skills}/orchestrate-agents/scripts/fleet.sh"   # shared store; stable across tools
"$fleet" doctor          # confirm tmux + which agent CLIs + agentspec + notifier are present

If doctor shows a CLI missing, it just drops out of the preference order.

Model

• Fleet = one tmux session. One per orchestration run.
• Window = a workstream group (per feature, per repo, per phase).
• Pane = one agent CLI. Reference agents by the pane id (%7) the helper

returns, never by name (the user's shell hook may auto-rename windows; pane ids and @fleet_* tags are immune).

• Guardian = a guardian subagent dedicated to one pane. Spawned at the same

moment the pane joins the fleet (at spawn, or at adopt for existing panes). The orchestrator never reads capture directly during the supervision loop; it reads the guardian's contract lines instead.

Tool preference when a workstream does not pin one: claude, codex first, then copilot, then agy. See [references/agent-clis.md](references/agent-clis.md) for each CLI's launch, headless, auto-approve, and resume flags.

Workflow

1. Survey first. "$fleet" doctor, then "$fleet" survey. Understand what

is already running before touching anything: every tmux session, window, and pane, which panes already run an agent CLI (AGENT=yes is a heuristic; capture the pane to confirm), and which carry @fleet_* tags from a previous run. Reconcile this picture with the request.

1. Plan around what exists. Turn the request into workstreams and agents,

preferring in order: (a) re-use a session that already matches the task (start on an existing session adopts it; adopt tags its agent panes), (b) extend that session with new groups/panes, (c) create a fresh fleet only when nothing suitable is running. State the plan to the user before mutating tmux: what you found, what you will adopt vs spawn, which tools. Never kill, restart, or repurpose a pane you did not create without the user's explicit consent.

1. Source prompts (optional). To assign a persona or skill as an agent's brief,

list what is available with agentspec manage list (add --format json to parse). Feed the chosen prompt to the agent in step 6.

1. Start or adopt the fleet: "$fleet" start . On an existing

session this adopts it (applies fleet display settings, emits EXISTING=1) instead of creating a duplicate. Print the ATTACH= line so the user can tmux attach and watch.

1. Adopt existing panes, then spawn only what is missing:

``bash "$fleet" adopt %3 --name api # existing pane joins the fleet as-is win=$("$fleet" group backend | awk -F= '/WINDOW/{print $2}') pane=$("$fleet" spawn "$win" claude --name api --dir ~/repo | awk -F= '/PANE/{print $2}') ` adopt tags a running pane (no restart) so list/state/ping and guardians cover it; tool is auto-detected from the pane's foreground command (override with --tool). spawn where tool is auto|claude|codex|copilot|agy`; it reuses the window's first pane for the first agent, splits for the rest.

1. Brief each agent. Wait until the pane shows a ready input box

("$fleet" capture until the CLI has booted), then send the task: ``bash "$fleet" send "Refactor the auth module. Run tests when done." ` send pastes multi-line text safely and submits. Do **not** re-brief an adopted pane that is already mid-task; capture` it, summarize what it is doing, and only send follow-up instructions the user has sanctioned.

6b. Dispatch a guardian for this pane. Immediately after the first successful send (or immediately after adopt, for panes already working), dispatch a guardian subagent for this pane (Claude Code: Task tool with subagent_type=guardian; Codex: /agent guardian inside the orchestrator session, or codex --profile guardian in a sidecar pane). Pass `, , the agent name, and the fleet.sh` path as the brief. The guardian owns this pane until terminal state. See Agent invocations below for the exact calls.

1. Open the dashboard (optional): "$fleet" dashboard turns the control

pane into a live, self-refreshing status table.

1. Supervise. The orchestrator no longer polls list/capture for liveness.

It consumes guardian contract lines (see Integration contract with guardian below) as events. The orchestrator's only loop responsibility is: receive contract line, decide if user input is needed, surface the line to chat, collect the user's reply, instruct the originating guardian to relay it via fleet.sh send.

1. Wrap up. Summarize each agent's outcome, citing transcripts (step on

"understand what's running"). Tear down with "$fleet" kill only after the user has what they need — and only if the fleet was created by this run; an adopted session is the user's, leave it attached.

Supervision loop (delegated to guardians)

Each guardian runs its own poll/classify cycle (see the guardian persona). The orchestrator treats each contract line as an event:

• needs-permission surface to chat, await user, relay decision through the

originating guardian.

• error surface to chat with the guardian's quoted error line, await user

instruction. Do not blindly retry auth or rate-limit errors.

• done mark the pane complete in the orchestrator's tally.
• stuck surface to chat and let the user choose between kill, send a nudge,

or wait.

• relayed confirmation only; advance to the next event.

fleet.sh list and fleet.sh capture remain available, but the guardian is their primary consumer. The orchestrator only reads them directly when a guardian is unreachable or the user explicitly overrides.

Tune idle sensitivity with FLEET_IDLE_SECS (default 30); the knob is read by each guardian, not the orchestrator.

Agent invocations

Spawn exactly one guardian per pane, immediately after step 6's first send (or immediately after adopt for panes that joined mid-task).

Claude Code (orchestrator inside Claude): use the Task tool.

Task({
  subagent_type: "guardian",
  description: "Guard pane %7 in fleet refactor-auth",
  prompt: "You are guarding pane %7 (agent: api, tool: claude) in fleet \
refactor-auth. fleet.sh is at ~/.agents/skills/orchestrate-agents/scripts/fleet.sh. \
Follow the guardian persona: poll list, capture on transition, emit one contract \
line per state change in the format GUARDIAN[%7]:  -  - . \
Never call send/spawn/kill/group on your own initiative."
})

Codex (orchestrator inside Codex): run inside the orchestrator session:

/agent guardian fleet=refactor-auth pane=%7 name=api tool=claude \
  fleet_sh=~/.agents/skills/orchestrate-agents/scripts/fleet.sh

or in a sidecar pane:

codex --profile guardian -- \
  "Guard pane %7 in fleet refactor-auth. fleet.sh=~/.agents/skills/orchestrate-agents/scripts/fleet.sh"

Both invocations brief the guardian with the four facts it needs: fleet name, pane id, agent display name, and the fleet.sh path.

Alerting (all four channels)

Guardians produce the alerts; the orchestrator is the megaphone. The in-chat channel is fed by the orchestrator forwarding the guardian's contract line verbatim, and the orchestrator must NOT paraphrase guardian output before showing it to the user.

"$fleet" ping "" [pane] fires every channel at once:

1. In-chat: the helper prints PING[...] to stderr. **You must also surface

the guardian's contract line, unmodified, to the user in this conversation** (the helper cannot write to chat itself). This is the channel that always matters.

1. macOS notification via osascript (Notification Center, with sound).
2. Terminal bell + status line on the pane that needs attention.
3. Status dashboard pane (when opened in step 7) reflects the new state.

Ping when, and only when: an agent needs permission, errored, finished a workstream, or is stuck. Do not ping for normal progress.

Understand what is running

Liveness comes from tmux (list, capture). Substance comes from agentspec, which reads each tool's own session store:

agentspec session list      # id | time | first prompt
agentspec session export  --last                   # full transcript as markdown

agentspec has no agy source yet (its gemini source reads the legacy Gemini CLI store, not ~/.gemini/antigravity-cli/). For agy panes, fall back to capture and resume with agy -c / agy --conversation .

Use the transcript to disambiguate an idle agent (done vs stuck), to summarize what each agent accomplished, and to brief a follow-up agent with prior context.

Safety (default: prompt on, ping to approve)

• Spawn agents bare so their normal permission prompts stay on. The orchestrator

detects a blocked agent and pings the user; the user decides.

• Do not pass --dangerously-skip-permissions, --yolo, --allow-all, or

codex --dangerously-bypass-... unless the user explicitly asks for an unattended / full-auto run for this fleet. Auto-approve flags per tool are in the matrix.

• Never approve a destructive or outward-facing action on an agent's behalf. Surface

it and wait.

• Survey before you mutate. Never spawn into, kill, restart, or repurpose a

pane or session you did not create this run; adopt it (with the user's intent clear) or leave it alone. kill is only for fleets this run created.

• One fleet per run; name it after the task, or adopt the session that already

hosts the work instead of starting a parallel one.

• Each pane has exactly one guardian. The orchestrator must not approve, kill,

or send to a pane without going through (or explicitly overriding with the user's consent) that pane's guardian. If a guardian crashes or its thread ends, treat the pane as unsupervised and ping the user before any further action on it.

Integration contract with guardian

Guardian-to-orchestrator messages are single lines, one per state transition, written to stdout:

GUARDIAN[]: - -

• ` is the tmux pane id (e.g. %7`), never the agent's display name.
• ` is exactly one of: running` (only on first attach),

needs-permission, error, idle, stuck, done, relayed.

• ` is a literal quote from fleet.sh capture` when

applicable, trimmed to fit one line; no paraphrase.

• ` is one of: awaiting user decision`,

awaiting orchestrator instruction, no action, relayed: "".

Orchestrator obligations:

• Surface every non-running line to chat verbatim.
• For needs-permission and stuck, block on user input before instructing

the guardian.

• Relay user decisions by telling the originating guardian to

fleet.sh send "" (guardian then emits a relayed line).

• Guardians never call send, spawn, kill, or group on their own

initiative; if a relay is needed, the orchestrator must instruct it.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: urmzd
• Source: urmzd/dotfiles
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.