Tostoevsky/TsienHsueShen

This repository hosts TsienHsueShen, a skill plugin built exclusively for DeepSeek Harness (DSH). It is a distilled version of the core methodology from Qian Xuesen and Song Jian’s revised edition of *Engineering Cybernetics*, stripped of domain-specific engineering examples to leave only transferable problem-solving frameworks. Once installed via the DSH CLI, a new skill named tsien-hsue-shen is added to DSH’s agent skill catalog. The agent can call this skill any time it is planning, executing, or reviewing any complex project to get structured engineering cybernetics-based methodology guidance.

Unlike domain-specific skills that add new knowledge to the agent, TsienHsueShen focuses on helping the agent reframe problems and approach complex work more systematically. When a project starts to spiral out of control, it does not just jump to asking what the next step should be. Instead, it prompts the agent to check core elements: whether the goal is clearly defined, whether the current project state is observable, whether key variables are controllable, if a working feedback loop exists, and when the project should stop instead of pushing ahead. Users do not need to learn any complex skill API to use it, and can simply invoke it with natural language prompts.

To install the plugin, you first need to have Node.js version 22.19 or higher installed on your system, and you must already have DeepSeek Harness set up and running. The plugin can be installed directly via the DSH CLI command, adding it from GitHub with no extra build step or compilation required. The plugin structure includes one prologue covering first principles and 13 topic chapters covering core methodologies like system perspective, problem framing, controllability analysis, and stopping rules. The code is released under the MIT license, and the skill content is a distilled re-creation that does not include large verbatim excerpts from the original book.