AI Agent Hub
Back to plugins
🖥️

dsh-subagent-response-promoter

Client Updated 2026.08.19

Run the following command in DeepSeek Harness:

dsh plugin install JachinShen/dsh-subagent-response-promoter

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

In DeepSeek Harness, install this plugin by running dsh plugin install JachinShen/dsh-subagent-response-promoter (source repository: https://github.com/JachinShen/dsh-subagent-response-promoter), then reinstall the corresponding DSH profile and restart the DSH process to activate it.

About this plugin

In multi-agent workflows a parent agent is often in the middle of a model inference or tool call when a child settles. The runtime parks the manager-authored subagent-settled response in the next safe step, but it may sit behind ordinary context, delaying the parent from handling the child callback. Similarly, a coordinator relay submitted via send_message cannot be read until the child finishes its current turn, introducing unnecessary latency. dsh-subagent-response-promoter is designed to eliminate that wait.

The plugin does two things. First, in the parent next safe step it stably promotes settled subagent responses ahead of ordinary context so the parent can consume child callbacks first without polling or interrupting an in-flight model or tool call. Second, it relocates the coordinator relay from the child next-turn inbox to its next-step inbox while preserving the original wake signal, letting a still-running child pick up the message at the nearest step boundary without waiting for the current turn to end or cancelling the active request.

It suits developers building background-subagent workflows on DeepSeek Harness who need low-latency deterministic message ordering between parent and child, especially when the parent is frequently busy and a polling approach would add unnecessary complexity.

Use Cases

  • A parent mid model call needs to process a settled child callback without polling or interruption
  • A coordinator relay must reach a still-running child at its nearest step boundary without cancelling an active request
  • Multi-agent pipelines require deterministic parent-child message ordering so subagent responses are not buried under ordinary context

Best For

  • Developers building background-subagent workflows on DeepSeek Harness
  • Agentic-system teams that require low-latency deterministic parent-child message ordering
  • Harness platform engineers who prefer a declarative approach over polling-based coordination