Preface¶

The plugin ecosystem of DeepSeek Harness (dsh) emphasizes “everything is a plugin.” The community directory is an independent site with no official affiliation to DeepSeek or Huafan, and should not be understood as an official app store.

For agent developers, a common pain point is not that the agent cannot write code or organize materials, but rather the lack of a stable method in complex projects: are goals clear, is status visible, can key variables be controlled, is feedback closed-loop, is uncertainty handled, and when should it stop. dsh-tsien-hsue-shen is a DSH skill plugin that provides methodological guidance for these types of issues.

What is This¶

Released by Tostoevsky under the MIT license, dsh-tsien-hsue-shen provides a DSH skill named tsien-hsue-shen to offer methodological guidance based on Qian Xuesen’s Engineering Cybernetics during model planning, execution, or project review.

The skill body of this plugin is derived from a learning distillation and recreation of Engineering Cybernetics (Revised Edition, 1980–1981, Science Press) by Qian Xuesen and Song Jian, and does not contain large excerpts from the original book. The original book is copyright of the authors and publisher.

Core Features¶

It provides methodological review, not an additional domain knowledge base. Verified capabilities include:

Provides the tsien-hsue-shen skill to DSH.
Provides methodological guidance during model planning, execution, or project review.
First checks goal definition, state observability, key variable controllability, feedback closed-loop, uncertainty, and stopping rules.
Helps the agent redefine the problem, identify leverage points, establish feedback, and determine when to stop.
The skill body contains 13 thematic chapters and a “First Principles” prologue, numbered 0–13.
The plugin is pure Markdown + minimal JavaScript, requires no build steps, and can be installed directly from GitHub.

Installation and Enablement¶

Installation requires the following Node.js version:

Node.js ≥ 22.19

First, start DeepSeek Harness:

npx @deepseek-ai/dsh web

This step is used to start the DSH web entry point.

Then, add the plugin to your profile:

dsh plugin --profile <your profile name> add github:Tostoevsky/TsienHsueShen

Here, <your profile name> needs to be replaced with the profile name you actually use. After installation, a skill named tsien-hsue-shen will appear in the session.

It should be noted that DeepSeek Harness is currently in the Developer Preview stage, and it is explicitly stated by the official team that there may be incompatible breaking changes. This plugin was tested based on the current dsh CLI plugin addition usage and the following startup method:

npx @deepseek-ai/dsh web

It will be continuously maintained as Harness updates.

Typical Usage¶

The following three are verified invocation examples. You can directly enter these tasks in the session.

Reviewing a research plan:

Use tsien-hsue-shen to review this research plan, first checking observability and controllability.

Handling non-convergent projects:

This project has gone through three iterations and still hasn't converged. Redefine the problem using Engineering Cybernetics methods.

Judging when to stop and deliver:

Do not continue optimizing the solution. First, determine if we have triggered a stopping rule, and what the minimum deliverable result is.

Suitable Scenarios and Notes¶

Suitable for:

Projects that iterate repeatedly but fail to converge.
Research plans or engineering proposals that need to check observability, controllability, and feedback closed-loops.
Need to judge whether to continue investing, and what the minimum deliverable result is.
Want to add a reusable project methodological framework to the agent.

Notes:

DSH plugins run with the permissions of the current dsh process; check source code and license before installing.
This plugin is under the MIT license; code can be reviewed on GitHub.
The skill body is a learning distillation and recreation, not containing large excerpts from the original book.
DSH is in Developer Preview stage, with possible breaking changes; after upgrading, it is recommended to reconfirm plugin availability.
The plugin directory is an independent community site, not DeepSeek or Huafan’s official app store.

Conclusion¶

The value of dsh-tsien-hsue-shen is not to answer specific professional questions for the agent, but to put goal, observation, control, feedback, uncertainty, and stopping rules into a single project review.

dsh-tsien-hsue-shen: Add Engineering Cybernetics-style project method to DSH