Introduction

In the context of DeepSeek Harness (DSH) plugins, if you wish to integrate a set of engineering workflows into an agent workflow, there are usually two approaches: scattering prompts throughout conversations, or maintaining scripts and templates yourself. dsh-mattpocock-skills provides a more centralized approach: it transplants the 25 skills from mattpocock/skills into a DSH plugin. After installing it to a specified profile, the model can load the corresponding skill content via the skill tool.

Below is an introduction to the positioning, features, installation methods, and typical usage of this plugin.

What is this

GongYuanCaiJi/dsh-mattpocock-skills is a DeepSeek Harness plugin transplanted from mattpocock/skills.

It contains 25 skills: 18 engineering-related and 7 productivity-related, such as tdd, code-review, to-spec, to-tickets, domain-modeling, grilling, etc. The skill content retains the original English text from upstream and is copied verbatim according to the “100% verbatim copy” rule.

This transplant only retains the 25 skills actually released by the upstream plugin. skills/misc, skills/in-progress, and skills/deprecated are outside the scope of this transplant.

Regarding licensing, this plugin uses the MIT License. The upstream mattpocock/skills is Copyright (c) 2026 Matt Pocock, while this transplant is Copyright (c) 2026 GongYuanCaiJi.

Core Features

Below are a few actionable points.

  1. On-demand loading. After installation, the model can see the skills catalog in the conversation. The catalog only exposes names and descriptions; the skill content is read only when the model actually invokes it.
  2. Support for explicit invocation. Users can load skills by starting input with /skill-name, for example /tdd, /grilling, /code-review.
  3. Word-for-word verification. The 76 skill files are byte-for-byte consistent with the upstream v1.2.3. The hash is stored in THIRD_PARTY_NOTICES.md, along with a directly copyable diff -rq verification command.
  4. Lightweight installation dependencies. This package has no build script and zero runtime dependencies; installing via git does not trigger the pnpm build-script whitelist. When installing from a local path, there is no need to run npm install first.

Installation

First, use the GitHub installation command to add the plugin to a specified profile:

dsh plugin --profile <name> add github:GongYuanCaiJi/dsh-mattpocock-skills

If installing from a local path, you can first clone the repository and then run:

git clone https://github.com/GongYuanCaiJi/dsh-mattpocock-skills.git
dsh plugin --profile <name> add ./dsh-mattpocock-skills

It declares engines.node >=20 in package.json. Regarding version numbers, the version in package.json is 0.1.0; in devDependencies, @deepseek-ai/cordis is 4.0.1, and @deepseek-ai/dsh-invariants, @deepseek-ai/dsh-llm, @deepseek-ai/dsh-scope, @deepseek-ai/dsh-skill are 0.1.0-rc.6.

Typical Usage

After installation, the model can see the skills catalog in the conversation. When a task matches a skill, the model uses the skill tool to load the full content.

It can also be invoked explicitly by the user, starting the input with the skill name:

/tdd
/grilling
/code-review

These names are examples found in the README and do not represent the complete list of skills.

Use Cases and Notes

This plugin is suitable for DSH users who want to reuse Matt Pocock’s engineering skills, especially those needing process-oriented skills like TDD, code review, spec/ticket workflows, domain modeling, or grilling, and who wish to verify that the transplanted content matches the upstream version.

Please note before use:

  • The plugin runs with the permissions of the current dsh process; please check the source code, license, and related files before installing.
  • This plugin uses the MIT License; the upstream skill content remains in English and is not translated.
  • It only includes the 25 skills actually released by the upstream plugin; skills/misc, skills/in-progress, and skills/deprecated have not been transplanted.
  • Skill invocation still relies on the model, tools, and permission settings within the current dsh session.

Conclusion

The value of dsh-mattpocock-skills is quite concrete: it integrates 25 engineering and productivity skills into the DSH plugin ecosystem, providing access points for on-demand loading, explicit invocation, and word-for-word verification.

GitHub:

https://github.com/GongYuanCaiJi/dsh-mattpocock-skills

Community Directory Page (Independent site, no official affiliation with DeepSeek / High-Flyer; refer to the actual page of the site):

https://www.skillhub.cn/plugins/GongYuanCaiJi/dsh-mattpocock-skills