AI Agent Hub
Back to plugins
🖥️

dsh-workspace-enhancement

Client Updated 2026.08.27

Run the following command in DeepSeek Harness:

dsh plugin install DobyChao/dsh-workspace-enhancement

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

In DeepSeek Harness, open the plugin marketplace and search for dsh-workspace-enhancement, or visit the source repository at https://github.com/DobyChao/dsh-workspace-enhancement for details, then run dsh plugin install DobyChao/dsh-workspace-enhancement to install.

About this plugin

In AI-assisted development, local directories and SSH remotes tend to live in separate windows. A single session is locked to one cwd, so flipping between reading a remote log, editing a local config, and firing a cross-machine command means context-switching between terminal and editor. dsh-workspace-enhancement collapses that loop: the session-header button lets you attach any number of side workspaces (a local folder or a directory on a remote host), each carrying its own read-only / read-write and exec on / off permissions that the model sees and acts on directly with no manual directory juggling.

The remote side rides a multi-hop ssh2 chain. ctx.subprocess and ctx.fs are swapped for transparent remote implementations, so bash, file I/O, PTY, and directory browsing all execute on the far end with zero changes to existing tool code. Cross-server execution is handled by sw_exec: pass a named server id and the command lands on that machine, with the target OS probed once per connection (bash -c on POSIX, pwsh -Command on Windows) and the result fed straight back into context. The machine registry, TOFU host keys, and OS-keychain passwords all stay in local ~/.dsh; nothing needs to be installed on the remote side, and the UI, prompts, and tool descriptions follow the language setting so a Chinese browser stays Chinese.

Built for full-stack engineers, SREs, and AI-assisted developers who juggle local code with several remote Linux or Windows servers. If your day involves jump chains, cross-host execution, remote file browsing, or you are simply tired of one-SSH-window-one-directory fragmentation, this plugin is worth a look.

Use Cases

  • Attach a local repo and a remote Linux directory to one session; the model reads and writes both with no manual directory switching
  • Reach an internal server through a multi-hop SSH chain with zero remote dependencies; bash, directory browsing, and PTY pass through transparently
  • Fire commands on multiple named servers via sw_exec; target OS is auto-detected and results flow straight back into the current session context

Best For

  • Full-stack engineers who hop between local code and several remote servers daily
  • SRE / DevOps engineers who rely on jump chains and cross-host execution
  • Developers who want AI to operate both local and remote codebases simultaneously