dsh-plugin-subagent-director
Run the following command in DeepSeek Harness:
dsh plugin install SeverusZh/dsh-plugin-subagent-director
Paste the following prompt into your AI chat to install this plugin:
Complete the plugin deployment in DeepSeek Harness by running the installation command in your terminal, and find the full open-source project source and development docs at https://github.com/SeverusZh/dsh-plugin-subagent-director.
About this plugin
Building complex multi-agent workflows in DeepSeek Harness often reveals a bottleneck: subagents rigidly inherit the parent’s LLM configuration, making it difficult to dynamically assign the right model for specialized tasks like code review or data processing. This rigidity leads to inefficient resource allocation and ignores the varying accuracy or latency requirements across different workflow stages.
Subagent Director solves this by introducing flexible role templates and dynamic model routing. You can define custom roles with distinct descriptions, injected personas, and dedicated provider/model bindings directly in the visual settings panel. The plugin enforces a smart four-layer resolution chain (call arguments → role binding → plugin default → parent inheritance) to auto-select the optimal LLM for every delegation. Coupled with hot-reload configuration, automatic main-agent task guidance, and extensible background continuability, it delivers granular control without touching core logic.
Ideal for developers, automation engineers, and teams orchestrating sophisticated AI pipelines, this plugin integrates seamlessly as a zero-intrusion layer. It empowers you to match exactly the right intelligence to each subtask, ensuring predictable performance, streamlined debugging, and highly adaptable multi-step reasoning environments.
Use Cases
- The main agent automatically delegates code review tasks to subagents bound to dedicated models based on preset guides.
- Switch subagent LLM providers on demand across conversation turns or complex pipelines to optimize inference costs.
- Quickly create translation and data analysis roles via the UI while independently specifying underlying reasoning models.
Best For
- Developers seeking granular control over inference latency and costs at different task nodes.
- Product teams aiming to rapidly build multi-step automation pipelines using a visual configuration interface.
- Engineers designing complex multi-agent architectures that require flexible and decoupled model routing.
Related Plugins
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.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.