In the plugin-based architecture of DeepSeek Harness (DSH), environment adaptation is a critical part of agent configuration. For Windows users, manually configuring a Minimal Mode Agent preset that works with PowerShell, Git Bash, or WSL2 often requires writing extra configuration files, which can be cumbersome and error-prone.
For this scenario, developer Dekurri provides the dsh-windows-minimal-adapter plugin. It is a workflow plugin designed to automatically generate a Minimal Mode Agent preset for Windows environments.
Core Features¶
The plugin provides the following capabilities:
- Automatic preset generation: After installation, it automatically generates a Windows-compatible Minimal Mode Agent preset.
- Multi-terminal support: Supports
pwsh(PowerShell), Git Bash, and WSL2. - Environment restoration: When the plugin is uninstalled, the generated preset is automatically removed to avoid leftover configuration.
Installation and Enablement¶
Add the plugin using the official command and specify a configuration profile (for example, demo):
dsh plugin --profile demo add github:Dekurri/dsh-windows-minimal-adapter
After installation, the preset configuration takes effect automatically.
Typical Usage¶
In addition to the command above, the plugin supports multiple installation methods. The following are reproducible examples:
# 使用本地目录
dsh plugin --profile demo add ./dsh-windows-minimal-adapter
# 使用本地 tarball 包
dsh plugin --profile demo add ./dsh-windows-minimal-adapter-0.2.1.tgz
# 使用 npm 包名
dsh plugin --profile demo add dsh-windows-minimal-adapter
Applicable Scenarios and Notes¶
- Suitable for users who need to quickly deploy DeepSeek Harness Agents in Windows environments.
- The plugin runs with the permissions of the current DSH process. It is recommended to review the source code and license before installation.
- The system must meet the Node.js version requirement (
>= 20).
This plugin simplifies development environment setup for Windows users by automating the configuration process. For more technical details and source code, please visit the project homepage.