DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
The dsh-chat-tools plugin provides conversation column width/position adjustment, numbered and shortcut-based question history, and Markdown outline directory functionality.
DSH Plugin Store is a native plugin marketplace for DeepSeek Harness that provides discovery, search, installation, and management of community plugins within the Harness interface.
The dsh-obsidian plugin integrates an Obsidian vault into DeepSeek Harness, offering 25 tools for memory management and file operations with secure, host-only execution.
dsh-clean-desktop-shell is a DSH plugin that adds a clean desktop shell to DSH Web, featuring system tray, single instance, and backend management without visual modifications.
A unified skills library for AI coding agents, offering domain and process skills compatible with tools like DeepSeek Harness and Claude Code.
dsh-cost plugin calculates DeepSeek conversation costs in real-time and displays a badge next to the input box, featuring session persistence and privacy-first design.
dsh-subagent-profile enables per-subtask control over subagent configurations, including preset, model, and reasoning effort, for optimized task dispatch and cost management.
An AI-powered PDF editing plugin that allows users to modify text in PDFs using natural language while preserving original layout, handling overflow, and supporting Chinese content.
dsh-usage is a usage query plugin for DeepSeek Harness that automatically identifies and displays the package/balance usage of various LLM providers in real-time, supporting zero-configuration and custom queries.
DSH File Viewer is a universal, read-only file preview layer for DeepSeek Harness that allows opening and inspecting files right inside the web UI without external applications.
A skin picker for DeepSeek Harness Web GUI with 25 skin families, supporting automatic switching between light and dark themes based on system preferences.
Generate several coding-agent patches, reject the ones that fail your tests, and let an LLM verifier rank the rest before you decide whether to apply the winner.