AI Agent Hub
Back to plugins
⚙️

dsh-cron

Workflow Updated 2026.09.03

Run the following command in DeepSeek Harness:

dsh plugin install fan56/dsh-cron

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install fan56/dsh-cron in your DeepSeek Harness to install this plugin; source code is available at https://github.com/fan56/dsh-cron .

About this plugin

In day-to-day DeepSeek Harness usage, you often want an agent to check CI status every ten minutes, generate a weekly report each Monday morning, or inspect disk usage hourly. The built-in dsh-schedule only offers session-scoped reminders and explicitly excludes cron expressions. dsh-cron fills that gap: profile-level recurring task scheduling with standard five-field cron expressions or fixed intervals, delivering prompts to whichever live root agent exists across TUI, Web, and Feishu.

A few design decisions keep it safe and predictable. Every recurring task must carry a validity window (capped at one year) and archives itself on expiry, eliminating infinite crons. The plugin owns the clock, so the model only expresses intent while all time math happens in the plugin. Delivery policy is either followup (queued, coalescing overdue fires) or steer (mid-turn injection for watchdog patterns); execution is either sub-agent (isolated worker with report backfill) or self (in-conversation, best-effort). Missed occurrences are skipped, never delivered late, preserving pure cron semantics.

Ideal for developers who need agents on recurring watch: CI monitoring, scheduled reports, disk inspections, health checks, or any check-it-at-time-X scenario. Just describe the need in natural language and the model handles the full lifecycle through cron_create, cron_list, cron_delete, and cron_report.

Use Cases

  • Periodically check CI build status and report results on schedule
  • Auto-generate weekly reports every workday morning and deliver to the team
  • Inspect disk usage hourly and inject an alert immediately when threshold is exceeded

Best For

  • Developers who need agents on recurring watch duties
  • Small teams using DeepSeek Harness multi-surface workflows
  • AI engineers who prefer natural language for scheduling tasks