triggers: block, or create one directly with reliant trigger create.
Declaring a schedule
skip records a skipped firing in the history instead of starting a second run on top of the first. A run that is still starting counts as going. A run that could not start does not hold the schedule up: the next fire runs in its place, and the one that failed to start is marked failed so it never runs late beside it. allow starts a concurrent run, which can pile up if each run takes longer than the gap between fires. The short catch-up window is deliberate: after an outage you get at most the last few minutes’ runs, not a replay of everything you missed.
From the CLI
--daemon (a machine id or hostname) and --message are required. --project defaults to the working directory, --workflow to your default workflow and --worktree (the workspace to run in; the flag keeps the git name) to the project’s main checkout. --preset, --param, --timezone, --overlap, --catchup-window and --disabled are optional. Every subcommand takes a trigger by id or by name.
Turning an automation on is yours to do, so create, update and enable refuse to run from an agent’s shell. An agent can still list, inspect, fire and disable them.
Manage them afterwards with:
fire skips the enabled and overlap checks. Deleting a trigger keeps its firing history; the runs it started still exist.
Related topics
- Automations overview
- No-machine runs
- Cloud machines & plans, for what happens when the machine is asleep at fire time