dsh-automation: A Scheduled Task Automation Plugin for DeepSeek Harness
titanwings/dsh-automation
Run coding tasks on schedule in fresh Agent sessions, manage schedules from DSH Web or an Agent.
Run coding tasks on a schedule in fresh Agent sessions, with schedules managed by users or Agents, leaving auditable run records.
dsh plugin --profile web add github:titanwings/dsh-automation- Category
- Workflow & Automation
- Platform
- DSH-Plugin
- Author
- titanwings
- Distribution
- Plugin
dsh-automation Key Features
dsh-automation Repository

dsh-automation Repo Summary
dsh-automation is a DeepSeek Harness (DSH) plugin that lets coding tasks run on a schedule in fresh Agent sessions, with schedules created and managed by users or Agents. It solves the problems of relying on old chat context, lacking explicit workspace and permission boundaries for unattended runs, and missing auditable run history. Core capabilities include creating schedule rules from DSH Web or any eligible root Agent, each dispatch starting a fresh root Agent and Session, and leaving durable run records. The plugin is maintained by titanwings, licensed under MIT, written in TypeScript, and last updated in 2026-08.
What core features does the dsh-automation DSH plugin support?
- One control plane, two ways in: Use the Automations conversation tab in DSH Web to create rules, pause/resume, run now, delete, and inspect recent runs; or ask any eligible root Agent in natural language. Six scoped tools let the Agent manage automations only for its exact workspace.
- Human-readable schedules: Create one-shot, fixed-interval, daily, or weekly rules. Daily and weekly schedules use an IANA time zone; the friendly form is normalized into a validated RFC 5545 RRULE for persistence and inspection.
- A clean execution boundary every time: Each dispatched occurrence receives a new Session ID and fresh root Agent, uses the saved prompt instead of source conversation history, captures workspace, cwd, Agent preset, model target, and permission preset, and includes an explicit
automationmessage source with automation ID, run ID, and scheduled time. The terminal result is derived from the actual DSH turn end, not merely "message delivered." - History that explains failure as well as success: Records definition revisions and durable run records, letting you inspect what ran, which revision it used, and how it ended.
How do I install dsh-automation?
Use the DSH plugin manager with the following command:
dsh plugin --profile web add github:titanwings/dsh-automation
After installation, you can create and manage automation tasks via the Automations tab in DSH Web or through natural language instructions to a root Agent.
How does dsh-automation differ from DSH Core Schedule?
DSH Core Schedule is for reminders in the current conversation, like "come back to this Session in ten minutes." dsh-automation handles a different job: "run this complete task independently every weekday and leave me a result I can inspect." The former returns to the same live Agent, while the latter starts a fresh root Agent and Session; the former takes a follow-up inside existing context, while the latter uses a saved, self-contained task; the former is scoped to the current Session Log, while the latter uses one canonical DSH workspace; the former history is conversation events, while the latter is definition revisions and durable run records.
What about the safety boundary of dsh-automation?
Each scheduled run executes within an explicit workspace and permission boundary, not inheriting implicit permissions from the source conversation. The plugin emphasizes that "a schedule is not permission"—scheduled tasks do not grant extra permissions but use captured permission presets. This ensures safety and auditability for unattended runs.
Which DeepSeek Harness versions and platforms does dsh-automation support?
The plugin targets the DeepSeek Harness framework, requires Node.js 22.19+, and is licensed under MIT. It is provided as a DSH bundle, integrated via the official plugin mechanism, and supports both DSH Web and root Agent management.