dsh-refactoring-guide
A DeepSeek Harness (DSH) plugin that provides the refactoring-guide skill for safe, incremental, behavior-preserving refactorings.
What it does
This plugin bundles the refactoring-guide skill, which applies safe, incremental, behavior-preserving refactorings using a test-protected, small-step workflow. The skill includes guidance for:
- Extract function/variable
- Rename
- Inline
- Decompose conditionals
- Remove duplication
- Replace conditional with polymorphism
- Introduce parameter object
The skill is automatically available in DSH sessions when this plugin is installed.
Installation
Installing from a local directory (recommended for development)
-
Navigate to the plugin directory:
cd E:\DeepSeek_Harness\plugins\dsh-refactoring-guide
-
Install into a DSH profile:
dsh plugin --profile web add .
Or from an absolute path:
dsh plugin --profile web add E:\DeepSeek_Harness\plugins\dsh-refactoring-guide
Installing from a published package (when available)
dsh plugin --profile web add dsh-refactoring-guide
Installing from a Git repository (when published)
dsh plugin --profile web add github:owner/dsh-refactoring-guide
Usage
No manual configuration is needed after installation. When the model encounters a refactoring-related trigger scenario during a session, it loads the matching skill automatically. You can also specify directly in conversation:
First give me an implementation plan for this refactor.
Or:
Refactor this function to improve readability.
How it works
The plugin registers a skill provider on ctx.skills that reads the bundled skills/refactoring-guide/SKILL.md file. The DSH skill system discovers the skill and makes it available to the model via the skill tool.
File structure
dsh-refactoring-guide/
├── package.json # declares dsh.bundle manifest
├── cordis.patch.yml # the layer applied when a profile lists this bundle
├── README.md # this file
├── lib/
│ └── index.js # plugin module that registers skill provider
└── skills/
└── refactoring-guide/
└── SKILL.md # the skill content
License
MIT