dsh-cron-scheduler

If you want your Agent to run scheduled tasks autonomously but refuse to tie every execution to the DSH Web process staying alive, dsh-cron-scheduler bridges that gap. Each schedule rule is stored as a standard Cron expression and handed off to the OS-level crontab, which spawns a headless session outside the DSH process. As long as the machine is on and cron is running, the task fires regardless of whether the Web UI is online. Late-night inspections, periodic reports, and data-sync jobs keep moving even when the Web server is briefly down.

On the management side the plugin exposes two entry points: a dedicated Settings page and in-conversation Agent tools. From the Settings page you can create, edit, pause, trigger, or delete rules just like any traditional cron job, with a live history panel showing queued, running, succeeded, failed, and skipped runs; clicking any entry opens the full session. In a chat you simply say "run XX every day at 9 am" and the Agent creates the rule through its automation tool chain, with the same list, history, update, and delete capabilities available. Cron syntax covers the full five-field form plus @reboot, @daily, @hourly, and other shortcuts; month and weekday names are accepted; timezone follows the system, matching classic cron behaviour.

Security and concurrency are handled out of the box. Every rule defaults to a read-only sandbox with an optional workspace-write tier, and no full-access mode is exposed. Concurrent triggers on the same rule are automatically marked as skipped to prevent duplicate execution. Session results use an eventual-consistency model: the Web process attaches sessions in real time when online, offline runs complete normally and are retroactively linked after a Web restart. Installation requires zero manual configuration, the plugin auto-resolves the dsh binary path for global installs, pnpm dev, and source checkouts, and a simple Web server restart plus a hard browser refresh is all that is needed.

This plugin is best suited for DeepSeek Harness users who want to turn repetitive, time-driven work such as ops inspections, scheduled reports, and batch data pipelines into fire-and-forget automations without spinning up a dedicated scheduling service or writing glue scripts.