Preface

DeepSeek Harness (DSH) adopts a plug-and-play architecture. When debugging agents or building complex workflows, prompt reuse often relies on manual copying and pasting. This approach is not only inefficient but also difficult to maintain. The dsh-prompt-library plugin aims to solve this problem by centrally managing reusable prompts and providing the ability to save, search, and edit them.

Plugin Positioning

This plugin is maintained by ManoloRemiddi and belongs to the memory plugin category. It allows you to save, search, preview, edit, and delete prompts in the DSH settings interface. Prompt text is rendered as plain text, and saving does not invoke a model or send chat messages. The plugin can run independently or connect to an existing Augmentor service.

Core Features

  1. Interface Management: Perform CRUD operations in Settings → Prompt library in DSH.
  2. Standalone Storage: Use a local SQLite database to store prompts.
  3. Service Integration: Connect to an existing Augmentor prompt service.
  4. Concurrency Handling: Handle concurrent save conflicts through a revision mechanism.

Installation and Activation

The plugin is distributed as a .tgz package.

  1. Download the Package
    Download the .tgz asset from the v0.2.1 release page.

  2. Run Installation
    Use the DSH plugin management command to install it. Assume that the web profile is used:

    dsh plugin --profile web add https://github.com/ManoloRemiddi/dsh-prompt-library/releases/download/v0.2.1/dsh-prompt-library-0.2.1.tgz
  1. Complete Configuration
    Installation requires a Linux environment, Node.js 22.19+, and an available DSH web profile. Standalone mode also requires Python 3. A standard Python build usually includes SQLite support.
    After installation, restart the DSH process and refresh the page. You can then use the plugin by entering Settings → Prompt library.

Usage

After installation, you can use the prompt library through the following steps.

Creating and Saving

On the settings page, click New prompt, enter the shortcut name and prompt text, and then click Save prompt.

Storage Modes

The plugin provides two storage modes:
* Standalone Mode (Default): Data is stored in $XDG_DATA_HOME/dsh-prompt-library/prompts.sqlite3. You can specify a different directory through the DSH_PROMPT_LIBRARY_DATA environment variable and specify a Python interpreter through AUGMENTOR_PYTHON.
* Augmentor Service Mode: Point the DSH_PROMPT_LIBRARY_SOCKET environment variable to an existing Augmentor Unix socket, such as $HOME/.local/state/augmentor/prompts.sock. The plugin connects to that service and does not need to start a Python process.

Inserting Prompts

Inserting a prompt into the chat editor requires a compatible Augmentor interface, such as a slash menu. The [clipboard] marker is stored as-is, while the specific expansion is handled by the consuming application.

Notes

  • Data Migration: Version 0.2.1 does not automatically migrate data from 0.1.0 or 0.2.0. A new standalone installation starts from an empty database, and the existing database remains unchanged. To preserve the data, you need to copy it manually or configure Augmentor service mode.
  • Concurrency Conflicts: If a concurrent save conflict occurs, the stale draft remains visible. When a conflict is detected, it is recommended to copy the content to a secure location, refresh, reselect the current prompt, and then apply your changes. The plugin does not automatically retry writes.
  • Storage Limits: Standalone mode has hard limits: a maximum of 1,000 prompts and a total content size not exceeding 500,000 bytes. Shortcut names support only 1–128 letters, digits, hyphens, or underscores.
  • Uninstallation Cleanup: Running dsh plugin --profile web remove dsh-prompt-library removes only the plugin interface. Saved database files and the Augmentor service remain unchanged and must be deleted manually.
  • Version Compatibility: The plugin has been verified on DSH 0.1.5-rc.1. Other versions may have different web/UI behavior.

Conclusion

dsh-prompt-library provides DSH with a lightweight local prompt management solution, suitable for workflows that require local management of a large number of prompts. For additional details or feedback, visit its GitHub repository or the SkillHub plugin directory.