dsh-cron

Usage

Adding a job by hand

Click + in the sidebar's Cron Jobs section header. The New job dialog configures everything on one screen:

• Name — lowercase, digits and -.
• Trigger — cron (5-field expression + IANA time zone), interval, or one-shot.
• Task — agent (a prompt executed unattended through the full dsh toolchain) or command (an argv to spawn).
• Preset / Access / Model — leave blank to inherit the host defaults.
• Working directory — type a path or browse via the folder icon.
• Timeout, On overlap, On misfire — see Features for the policy semantics.

Create & enable persists the job (a "manual" chip marks it apart from config-declared jobs). Afterwards, each row's ⋯ menu offers Run now / Pause schedule / Edit job / Delete job; clicking a row expands its run history, and clicking an agent run opens that run's full session replay.

Adding a job from a session

Just ask the agent in any session:

Every Monday at 07:00 review our outdated dependencies and save an upgrade checklist to reports/deps-audit.md.

The bundled cron-create skill (auto-registered when the host has a skill registry) walks the model through collect → confirm → create → verify, calling the cron_create tool under the hood; cron_delete removes a manual job the same way. Jobs created from a session are ordinary manual jobs — the exact same overlay the web dialog writes — so they show up in the sidebar immediately and can be edited there later. The read/steer tools (cron_list, cron_runs, cron_run_now, cron_enable, cron_disable) work on config-declared jobs too.