Status

/onex:status — Foreground Status Routing

Skill ID: onex:status Version: 1.0.1

Announce at start: "I'm using the status skill."

What this skill does

Delivers a PR / merge-queue status snapshot to the foreground session without inline repository polling. Foreground sessions must route state requests through the merge-sweep pipeline's authoritative view.

Routing decision:

1. Call TaskList to enumerate active tasks in the current team.
2. If a task named merge-sweep-lead exists with status in_progress:
   • Send {"type": "push_snapshot"} via SendMessage(to="merge-sweep-lead").
   • Wait for the teammate to reply with its current snapshot, then surface that reply to the user verbatim. Do not post-process or re-query GitHub.
3. If no merge-sweep-lead task is found:
   • Spawn a one-shot snapshot agent (see One-Shot Snapshot Agent below).
   • Surface the agent's returned snapshot to the user once it completes.

Under no circumstances should this skill or any agent it spawns query PR state directly from the foreground session.

Routing decision pseudocode

tasks = TaskList()
lead = first task in tasks where name == "merge-sweep-lead" and status == "in_progress"

if lead:
    SendMessage(to="merge-sweep-lead", message='{"type": "push_snapshot"}')
    # teammate replies with snapshot JSON; surface to user
else:
    Agent(
        name="status-snapshot",
        prompt=ONE_SHOT_SNAPSHOT_PROMPT,
        model="claude-haiku-4-5-20251001",
    )
    # surface agent output to user

One-Shot Snapshot Agent

When no merge-sweep-lead is running, spawn a minimal snapshot agent with this prompt (inline, no separate file):

You are a one-shot status snapshot agent. Your only job:

1. Run this exact standalone-CLI preflight and dispatch from the current directory:

<!-- status-snapshot-command:start -->
```sh
if ! command -v onex >/dev/null 2>&1; then
  printf '%s\n' "ERROR: Required standalone ONEX CLI was not found on PATH. Install it with: pipx install 'omnibase-core>=0.39.0'" >&2
  exit 127
fi
exec onex run-node node_pr_lifecycle_orchestrator --input '{"dry_run": true, "inventory_only": true}'
```
<!-- status-snapshot-command:end -->

2. If the command exits nonzero, send its error output verbatim to "team-lead" and stop. Do not install anything, search repository virtualenvs, or query GitHub directly.
3. On success, parse the returned ModelPrLifecycleResult JSON.
4. Emit a concise snapshot: open PRs, merge-queue state, CI failures, blocked tickets.
5. Send the snapshot text to "team-lead" via SendMessage.
6. Stop immediately after sending.

Never query repository state directly. Never open PRs or push code. Read-only
snapshot only.

Use model="claude-haiku-4-5-20251001" for the snapshot agent — this is a read-only reporting task, not a reasoning task.

Failure modes

Why this routing exists

Foreground PR/merge-queue polling is a recurring source of stale reads, rate-limit waste, and inconsistent snapshots. The foreground session has no cache of the merge-sweep lead's view, and the sweep lead may act on a newer snapshot milliseconds later.

Routing through merge-sweep-lead when it exists gives the foreground session the authoritative view that the sweep lead is already maintaining.

See Also

• /onex:merge_sweep — launches or resumes the merge-sweep pipeline
• /onex:system_status — full platform health (infra, Kafka, DB); not PR state
• omnimarket PR lifecycle orchestrator — backing node for PR inventory