AI Agent Hub
Back to plugins
🖥️

dsh-insight-dock

Client Updated 2026.09.16

Run the following command in DeepSeek Harness:

dsh plugin install POPCORNBOOM/dsh-insight-dock

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

Run dsh plugin install POPCORNBOOM/dsh-insight-dock to install this plugin in DeepSeek Harness; the source repository is at https://github.com/POPCORNBOOM/dsh-insight-dock

About this plugin

While an agent is deep in a task, a side thought keeps surfacing: "wait, that other thing could…" Interrupting the current flow to address it is costly, but dropping it entirely means losing it forever. dsh-insight-dock gives the agent a strictly bounded parking spot for exactly that kind of observation—a short note that floats above the composer until you decide to explore it or dismiss it.

Every note carries a context anchor: the session sequence number at the moment it was created. When you choose to hear it out, insight_load reads back the real conversation log around that anchor from the durable session record, so the original context is recovered even if the agent's working memory has since been compacted. The note is the ticket; the log is the library.

Notes live in memory only and expire after a configurable lifetime (default five minutes). Every ending path deletes the record—there is no archive, no disk write, no persistence to clean up. Prompt strength runs from fully silent to an active nudge for the agent to scan before starting a turn, so you tune how eagerly the dock gets used. If you work with the DSH Web GUI and want a lightweight "second sight" channel that never derails the primary task, this is the plugin to reach for.

Use Cases

  • Park a bounded side note mid-task without derailing the primary flow
  • Read back the anchored log window from the durable session record
  • Tune prompt strength to control how eagerly the agent records insights

Best For

  • Engineers driving long-running tasks in the DSH Web GUI
  • Teams that want a second-sight channel without agent distraction
  • Users who prefer a lightweight, in-memory, self-expiring side-note mechanism