AI Agent Hub
Back to plugins
🤖

dsh-subprocess-win32

Model Inference Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install delightedMaster/dsh-subprocess-win32

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install delightedMaster/dsh-subprocess-win32 in the DeepSeek Harness terminal to install this plugin; source available at https://github.com/delightedMaster/dsh-subprocess-win32

About this plugin

DeepSeek Harness ships Unix-flavored process handling by default, which on Windows tends to leak orphaned processes, drop signals, and stall terminals. dsh-subprocess-win32 replaces that path with a native Windows process-tree inspector and clean termination, and bundles a one-command PowerShell lifecycle manager (doctor, setup, update, rollback, uninstall) so you get a stable runtime with zero orphaned processes and no silent PATH tampering.

Two Agent presets are included and rendered automatically by the lifecycle manager: Minimal (Windows) keeps token overhead to a minimum with a persistent Git Bash session and str_replace_editor, ideal for everyday coding; Anchored Standard (Windows) uses two-phase progressive loading, starting with zero tool bloat and unlocking PowerShell, Web, and Subagent tools on demand, suited to heavier task pipelines. No hand-edited YAML required.

The bundle patch only overrides the default subprocess row on Windows; Linux and macOS behavior stays untouched. It is aimed at developers running DSH on Windows x64 and engineering leads who need a consistent, leak-free Windows development environment across their team.

Use Cases

  • Fixing DSH terminal process leaks and signal drops on Windows
  • Daily coding with a low-token lightweight Agent preset
  • Standardizing Windows dev environments across a team with zero orphaned processes

Best For

  • Developers using DSH on Windows x64
  • DevOps engineers needing stable subprocess management and lifecycle tooling
  • Engineering leads standardizing DSH Windows environments across a team