| name | waiting-for-campaigns |
| description | Use when a RoboVAST campaign has been started and the turn is about to end — how to wait for it without holding the conversation open, and what to say if you are not going to wait. |
Waiting for a campaign
start_campaign returns as soon as the campaign is named. The campaign then runs for
minutes, hours or days. Reading one status and reporting the result is the mistake this
skill exists to prevent: it tells the user a campaign finished when it has barely begun.
Wait for it, in the background
Bash(command="vast campaign wait <campaign_id>", run_in_background=true)
Run it as the whole command — nothing chained after it. The exit code is the entire
signal, and ; echo … makes the harness observe the wrapper's status instead: a failed
campaign then reports success.
| exit | means |
|---|
| 0 | finished, past postprocessing |
| 1 | failed or stopped |
| 2 | --timeout elapsed; the campaign is still running |
| 3 | the service has no phase for that id — a typo, or it died before recording one |
| 4 | stalled — nothing completed for longer than one run may take |
| 5 | a running job's simulator reported a fault (sim time not advancing, and the like) |
Backgrounded, it costs you nothing: you stay free, and you are notified when it exits.
4 and 5 are a hand-off, not an ending
The campaign is still running and nothing is waiting on it now. Neither exit touched the
run. This is the waiter telling you something is wrong so you do not have to remember to look
— which is the whole reason it exits at all, since a wedged campaign holds running for its
whole life and would otherwise never end the wait.
The message names the diagnosis and the next call. Work through it cheapest-first —
get_job_state, then get_job_log(summarize=True), then reproduce in a copy with
exec_in_container, and only then exec_in_job, which is recorded against the run. Then
settle the campaign again: background vast campaign wait <campaign_id> (it will not exit on the
same thing twice, so resuming works), or stop_campaign.
Or say you are not waiting
A sweep can legitimately run for days, and no in-session wait survives a closed session.
Not waiting is fine — silently not waiting is not. Say which campaign you are leaving,
and that ntfy announces the end.
Or abandon it
stop_campaign(campaign_id=…) if it should not continue. That settles it too.
What the hook does
The first attempt to end a turn on a campaign nobody is waiting for is blocked, once, and
then allowed. It is a floor under the three choices above, not a fourth: it never waits
for you, and it never blocks a second time for the same campaign.
