AI Agent Hub
Back to plugins
🖥️

dsh-any-attachment

Client Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install Zenjibad/dsh-any-attachment

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

To install the dsh-any-attachment plugin, please execute `dsh plugin install Zenjibad/dsh-any-attachment` in your terminal. Source: https://github.com/Zenjibad/dsh-any-attachment

About this plugin

When handling complex tasks with DeepSeek Harness, seamlessly passing local files or workspace files to an AI agent can often feel cumbersome. The dsh-any-attachment plugin aims to revolutionize this experience. It allows users to attach any type of file directly from the Web UI—whether code snippets, documents, or images—without the need to manually copy paths or worry about files cluttering up your workspace.

The plugin's core innovation is "pathless mentions." When you add an attachment by typing @ or dragging and dropping a file, the system generates a simple @name tag in the composer. The referenced file is securely stored in a private directory ($DSH_HOME/attachments-any/) without polluting your workspace. The agent then resolves these tags and extracts the exact absolute path in the sent message, reading the file directly using its native tools.

Whether you are a developer needing to quickly share code or a researcher needing to upload documents, this plugin significantly boosts efficiency. It fully supports the built-in image pipeline for rasters and ensures file privacy, making it the ideal choice for users looking to maximize an AI agent's context capabilities without sacrificing workspace cleanliness.

Use Cases

  • Quickly upload code files for code review or debugging.
  • Upload documents to get detailed analysis or summary.
  • Process images for visual recognition or design analysis.

Best For

  • Developers needing to pass files to agents to enhance capabilities.
  • Data scientists who wish to keep their workspace tidy.
  • AI researchers automating tasks using dsh workflows.