Introduction

Research tasks are the easiest scenario where agents can appear “correct but are unusable”: searching for information without sources, proposing hypotheses without drawing evidence boundaries, experimental plans that cannot be executed in the lab, and figures that fail to meet journal publication requirements. For those conducting literature reviews, method validation, or responding to reviews, these constraints are indispensable.

The DSH philosophy is “everything is a plugin,” allowing domain workflows to be packaged as reusable skills without rewriting specifications in the prompt every time. The dsh-scientific-research introduced below is such a skill package for research scenarios.

First, a note on the information source: the GitHub README was not successfully scraped, so the functional descriptions below are based on the summary provided by the community directory page and have not been cross-checked with the repository content. Please refer to the actual content of the repository before installation.

What is it

First, the basic information:

  • Name: satan9394/dsh-scientific-research
  • Maintainer: satan9394
  • Category: Workflow
  • Repository: https://github.com/satan9394/dsh-scientific-research

According to the summary on the directory page, it covers major research stages: from database queries, literature reviews, and hypothesis generation to experimental plans, peer review responses, method validation, and figure generation. The project is inspired by K-Dense-AI/scientific-agent-skills (33.9k★).

Capabilities Listed on the Directory Page

The summary lists six capabilities; here is a detailed explanation of each:

  1. 100+ database queries with sources. The core requirement for research retrieval is traceability; every conclusion must be traceable to its source.

  2. Literature reviews and evidence-bounded hypotheses. Hypothesis generation is limited to the scope of existing evidence, rather than allowing the model to freely generate content.

  3. Experimental plans and budgets. Output the plan while providing a budget estimate.

  4. Peer review responses. Assistance in organizing responses to reviewer comments.

  5. Method validation, referring to ICH, USP, and CLSI. These three respectively correspond to the regulatory systems of the International Council for Harmonisation, the United States Pharmacopeia, and the Clinical and Laboratory Standards Institute, which are common reference standards for method validation roles.

  6. Publication-grade figures. Output figures according to publication requirements.

Note: The descriptions above are from the directory page summary; specific implementations and triggering methods for each capability need to be confirmed in the repository.

Installation and Usage

Neither the directory page nor the scraped materials provide verifiable installation commands or license information. We will not speculate or concatenate commands here. Before installing, do three things:

  1. Open the repository, read the README and source code, and confirm the skill organization and triggering methods;
  2. Confirm if the license meets your usage scenario;
  3. After confirming everything is correct, install according to the repository instructions.

Repository and directory page addresses:

https://github.com/satan9394/dsh-scientific-research
https://www.skillhub.cn/plugins/satan9394/dsh-scientific-research

Note that the directory page is an independently maintained community site and has no official affiliation with DeepSeek or Hanhua.

Typical Usage

There are no reproducible usage examples in the scraped materials, so we will not fabricate any here. The least costly way to verify is: after installation, let it run a minimal task first, such as a single database query with sources, and manually check if the sources are real and traceable. After confirming reliability, incorporate it into the formal research workflow and expand its usage.

Use Cases and Notes

Suitable for:

  • Researchers who need to conduct literature reviews, write proposals, or apply for grants;
  • Method development and validation professionals whose work revolves around ICH, USP, and CLSI;
  • Paper authors who need to draft peer review responses;
  • Developers who want to integrate research workflows into DSH agents.

Notes:

  • The plugin runs with the permissions of the current DSH process; check the source code and license before installation;
  • The license for this plugin has not yet appeared in the verifiable materials; you need to confirm it yourself at the repository before use;
  • The functional summary comes from the directory page clues; actual capabilities should be based on the repository.

Conclusion

In a nutshell: dsh-scientific-research packages tasks in research scenarios that require the highest level of traceability—database queries with sources, evidence-bounded hypotheses, method validation, and publication-grade figures—into a single DSH skill package. Whether it is good to use is for you to judge after checking the source code in the repository.

  • Directory page: https://www.skillhub.cn/plugins/satan9394/dsh-scientific-research
  • GitHub: https://github.com/satan9394/dsh-scientific-research