dsh-delayed-task
Run the following command in DeepSeek Harness:
dsh plugin install m-guo-2/dsh-delayed-task
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install m-guo-2/dsh-delayed-task in your DeepSeek Harness terminal to install the plugin; the full source repository is at https://github.com/m-guo-2/dsh-delayed-task
About this plugin
Agents frequently hit the “I cannot decide right now, let me check again in a few minutes” situation. dsh-delayed-task lets an Agent genuinely return to the original session when the scheduled moment arrives, gather fresh evidence, and re-evaluate the decision — rather than blindly executing a plan written two minutes ago or merely firing a system notification.
When creating a delayed task, the Agent records an objective decision context: what it observed, why it chose to defer judgment, and what needs to be re-confirmed at the due time — not a pre-scripted future command. At the scheduled moment, the DSH background service resumes the original session via agents.resume and injects a traceable delayed-task event. The Agent then assesses whether the old plan still holds under current conditions. The plugin offers two authorization tiers: the default reassess-only mode permits only read and search tools for gathering evidence, with no writes, deletions, or network calls; users may also explicitly grant a bounded grant that fixes an exact tool allowlist, a maximum call count, and an expiry window, and the Agent executes only within those bounds at due time. Closing the browser tab or letting the session go cold does not prevent triggering; as long as the DSH background process is alive, the task wakes the Agent on schedule. If the process or machine stops and later recovers, overdue tasks are scanned and retroactively triggered.
It is built for DSH users who construct “check back later” workflows: polling deployment health a few minutes after launch, waiting for CI pipelines or async data jobs to reach the next state, deciding whether to retry after a rate-limit or transient failure, continuously observing a condition while letting the Agent dynamically adjust the next check interval, and keeping results pinned to the original session even after the browser tab is closed. If you need a fixed command at a fixed time, cron is the right tool; if you need to wake an Agent at a future decision point to think again, this plugin is designed for exactly that.
Use Cases
- Polling deployment health a few minutes after launch without blocking the conversation
- Reassessing next steps once a CI pipeline or async job reaches its next state
- Deciding whether to retry after a rate limit or transient failure based on fresh evidence
Best For
- Developers building DSH Agent workflows that need cross-time decision review
- Teams wanting Agent results to persist in the original session even after the tab closes
- Operators who need dynamic retry decisions after rate limits or transient failures
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.