AI Agent Hub
Back to plugins
dsh-attention preview

dsh-attention

Workflow Updated 2026.08.30

Run the following command in DeepSeek Harness:

dsh plugin install zhaozixi/dsh-attention

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

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

About this plugin

A long goal loop, a batch of background jobs, an Agent reasoning pass—any of these can leave you staring at a progress bar for minutes or even hours. dsh-attention turns that idle wait into productive micro-moments: five seconds after a task starts, a floating panel pops up with a 30-second to 3-minute activity suggestion (a quick stretch, tidying your desk, five vocabulary words). Click Done and the next one appears immediately; when the task finishes, the panel collapses and generates a utilization report so you can see exactly how much downtime you actually used.

The plugin watches four signal sources—agent status, goal rounds, background jobs, and task-board polling—to detect task start and end windows, so reminders only fire while work is genuinely in progress. It ships with 43 preset micro-activities across four categories, shuffled per session with no repeats. You can add, edit, or remove activities through the built-in pool editor in the sidebar, and the engine picks up the new list the moment you save. After a DSH restart, in-flight task windows are automatically re-seeded so no state is lost. The floating panel sits on the highest z-index with a live display of uptime, suggestion count, and countdown to the next nudge.

Ideal for developers who regularly run goal loops, Agent inferences, task-board pipelines, or batch background jobs in DSH and feel the drain of idle staring-back. The plugin only reads the local event bus and schedules timers; it makes no network calls, reads no workspace files, and all HTTP endpoints bind to 127.0.0.1 only—zero extra attack surface to worry about.

Screenshots

Use Cases

  • Stay engaged during DSH goal loops or Agent inferences with timed micro-activity nudges
  • Keep pace with batch background jobs by clicking Done to instantly advance to the next suggestion
  • Automatically re-seed in-flight task windows after a DSH restart so no reminder state is lost

Best For

  • Developers who regularly run long DSH tasks
  • Efficiency-minded users who want to convert idle wait into productive micro-moments
  • Agent-workflow practitioners focused on attention management and pacing