Introduction

The design philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” In practical development, maintaining a set of standardized SKILL.md files often requires manual copying, which is not only inefficient but also makes it difficult to ensure version consistency. The dsh-xingtu-skills plugin packages 26 engineering-verified, production-grade AI agent skills into a standard DSH plugin, aiming to address the standardization of skill distribution and management.

Plugin Overview

This plugin is maintained by the user xingtu1996 and is categorized under “Model Inference.” It bundles the XingTu skill set, providing a complete set of capabilities ranging from token optimization and code debt management to engineering practices and content review.

Core Features

The plugin contains four skill modules:

  • Token Compression: The Caveman series, focused on compressing input and output tokens to optimize model interaction efficiency.
  • Minimalism and Debt Management: The Ponytail series, used to identify and manage debt in code and architecture while keeping systems clean.
  • Engineering Practices: Covers development workflows such as safe refactoring, migration, surgical patching, verification, and stopping.
  • Content and Media: For XingTu creator content scenarios, providing capabilities such as adversarial review, removing AI-like tone, and fact-checking.

Installation and Activation

Use the officially recommended npm command to install:

dsh plugin add dsh-xingtu-skills

After installation, DSH’s session skill directory automatically loads all skills. The system uses a progressive disclosure mechanism, so there is no need to manually copy files; skills are triggered on demand when needed.

Technical Details

  • Dependencies: The plugin is designed to be zero-dependency. No copy of Cordis is bundled inside the package, and @deepseek-ai/dsh exists only as an optional peerDependency.
  • Verification Process: The plugin has passed dsh-plugin-dev check (static checking) and dsh-plugin-dev verify (installation, startup, and uninstallation verification in a clean environment).
  • Directory Structure: Standard plugin structure, including the lib/index.js entry and the cordis.patch.yml declaration file.

Use Cases and Notes

This plugin is suitable for DSH developers who need standardized skill management. The plugin runs with the privileges of the current DSH process. Before installing, it is recommended to review the source code and license (MIT License).

Summary

dsh-xingtu-skills provides a complete set of engineering-oriented skills. It can be integrated with the dsh plugin add command. For more details, see the GitHub repository or the DSH ecosystem directory.