dsh-email-push-master: An Offline Email Reminder Plugin for the DSH Agent
dsh-email-push-master is a DSH plugin designed to address missed progress in long-running tasks. It sends one-way email notifications to users, supporting triggers when tasks complete, become blocked, or require decisions. With zero core dependencies, it uses Node.js built-in modules to avoid Windows port 465 compatibility issues, and includes intelligent error handling, automatic SMTP host inference, and self-check functionality. It is easy to install, secure to configure, and compatible with multiple AI coding agents.
Read MoreIntroducing the DeepSeek Harness Hosted Approval Plugin
DeepSeek Harness (DSH) relies on manual approval for its native permission controls, and the `dsh-managed-approval` plugin provides a solution. It introduces Codex-like managed approval logic, adding a session-level **"Approve for me"** preset to the DSH permission menu. The plugin grants one-time permissions for operations through risk and authorization assessment, supports MCP-aware review, and falls back to manual approval upon failure. The plugin is in beta, requires DSH 0.1.0-rc.6 to be pinned, and is suitable for developers who want to reduce manual intervention.
Read Moredsh-model-balance: Automatic Detection and Query of DSH Input Line Balances
The dsh-model-balance plugin is designed to eliminate the workflow interruption caused by manually checking balances in DeepSeek Harness (DSH). It automatically identifies model providers and displays a balance or quota icon directly in the input line. The plugin supports mainstream service providers such as DeepSeek and Moonshot, as well as custom endpoint configurations, and features automatic detection, hover-to-view, and secure host-side key handling. Users can install it with a simple command and monitor resource usage in real time in DSH Web, thereby improving development efficiency.
Read Moredsh-plugin-tree: DSH Plugin Dependency Layered View
dsh-plugin-tree is a plugin for the DSH web client that upgrades the "Settings -> Plugins" page into a visual dependency management tool. It provides a hierarchical plugin tree, a dependency graph, and node details, supporting search-based filtering, enabling or disabling plugins, and checking for npm updates. This tool requires no configuration; it manages plugin dependencies by reading loader information and writing configuration files, helping developers intuitively organize complex plugin stacking structures.
Read Moredsh-mobile-ui: Mobile Responsive Overlay for the DSH Web Settings Panel
dsh-mobile-ui is a responsive CSS override plugin maintained by jackuh105, designed to fix mobile layout issues caused by DeepSeek Harness's default 800px width. When the viewport width is 680px or less, the plugin converts modals into full-screen vertical forms, adjusts the navigation bar into a top tab bar that scrolls horizontally, and optimizes scrolling and safe-area handling in the content area. Installation requires specifying the workspace root directory with the `-w` parameter and completing the process using the `dsh plugin add` command; it takes effect after restarting the web profile. Developers can edit the breakpoints in `lib/index.js` or adjust CSS selectors for different versions.
Read Moredsh-skill-manager: DeepSeek Harness Web-based Skill Management
`dsh-skill-manager` is a web-based GUI plugin for DeepSeek Harness (DSH), designed to address the lack of intuitive in-browser skill management. It allows users to browse directories, preview content, import, create, and delete skills through the web interface, with import support for GitHub and local sources. Once installed, users can operate it via the `/api/dsh-skill-manager/*` API in the settings panel. Notably, the plugin only allows deletion of skills located within a specific directory and supports symbolic link operations, enabling it to work around path resolution issues caused by a mismatch between the working directory and the Git root directory, thereby improving the efficiency of skill management in web-based environments.
Read More@richie.liu/dsh-hybrid-coder: Dual-Model Routing Strategy Plugin
The `@richie.liu/dsh-hybrid-coder` plugin aims to resolve the conflict between resource waste and performance limitations in DeepSeek Harness (DSH) agents. It rewrites request targets to enable dual-model routing: a premium model handles planning and emergencies, while a local small model (e.g., Ollama) executes ordinary steps. Core features include plan-mode upgrades, automatic fallback on consecutive failures, request-level fault recovery, and degradation mechanisms. The plugin is currently experimental and requires specifying premium and local routing parameters in the configuration file.
Read Moredsh-file-picker: Native Windows File Picker Plugin
This plugin is maintained by JackeyWilder and is designed to address the need for DeepSeek Harness (DSH) agents to directly access the local file system. It provides a native Windows file picker, allowing users to select a file and inject its path into a draft in the format `@path:`. Since no official installation command is available, users need to obtain the source code from the GitHub repository. Given that the plugin relies on local system interactions and has an unclear origin, users are advised to inspect the source code themselves to confirm its security and licensing terms.
Read Moredsh-telegram-control: Telegram Remote Control Plugin for DeepSeek Harness
dsh-telegram-control is a client plugin for DeepSeek Harness that enables users to remotely control local DSH processes through a Telegram bot. Built into the Harness framework, it requires no external runtime and supports command- or inline-button-based management of sessions, background tasks, real-time output delivery, and permission approvals. Installation requires configuring a token and an allowlist, as well as meeting the Node.js 22 and DSH runtime requirements, ensuring secure and efficient maintenance of complex agent workflows on mobile devices.
Read MoreDeepSeek Harness Plugin Introduction: dsh-minimax-image
`dsh-minimax-image` is a plugin for DeepSeek Harness (DSH) that generates images directly within a conversation and saves them locally. It is based on the MiniMax `image-01` model, supports multiple aspect ratios, and integrates tool calling, credential reuse, and automatic saving. During installation, the plugin source must be registered in DSH's `web` and `headless` profiles, and parameters such as the API key must be configured via environment variables or a configuration file. Generated images are saved in the `generated/` directory of the workspace. When in use, the model automatically invokes the `image-gen` tool, so users only need to describe the desired image in the conversation. This plugin is suitable for scenarios where image creation needs to be integrated into the conversational flow.
Read Moredsh-subagent-response-promoter: DeepSeek Harness Subagent Response Promoter Plugin
`dsh-subagent-response-promoter` is a DeepSeek Harness plugin designed to resolve delayed processing of subagent responses while the parent agent is busy. Using a host-side architecture, it prioritizes subagent responses and coordinator relays ahead of ordinary context at agent step boundaries, ensuring the parent responds to subagents first without interrupting ongoing model reasoning or tool calls. It is simple to configure and suitable for complex agent workflows that require stable response priorities.
Read Moredsh-token-stats: Token Usage Statistics Plugin for DeepSeek Harness
dsh-token-stats is a local plugin for DeepSeek Harness (DSH) designed to generate visualized token statistics reports by reading session logs. The plugin supports aggregating data by hour, day, week, and month, displays line charts for input, output, and cached tokens, and provides cache hit rate analysis. Data is stored locally without cloud transmission, and an HTTP API is provided for queries. Installation requires the `dsh web` profile, and the configuration file is located in `cordis.patch.yml`.
Read Moredsh-htmlrun: A DeepSeek Harness Plugin for Publishing Local HTML as html.run Links
`dsh-htmlrun` is a DSH model tool plugin designed to solve the problem that sandbox environments cannot directly call the html.run publishing API. By wrapping a local CLI tool, it integrates the capabilities of publishing, updating, and checking the status of static HTML files into model tools. The plugin does not require additional key configuration; it leverages html.run's local authentication mechanism and supports commands such as `check` and `publish`. A model can directly invoke this tool to convert local HTML into shareable links, making it suitable for quickly publishing single-file static pages.
Read Moredsh-ui-task-notify: Web Task Completion Notification Plugin
`dsh-ui-task-notify` is a DeepSeek Harness (DSH) plugin designed to prevent easily missed task-completion and pending-interaction signals when work is running in the background or the page is minimized. It leverages the browser Notification API to trigger system notifications and sounds when an agent finishes work or awaits human intervention (such as approval), without requiring extra UI or RPC. The plugin supports configuration options such as notifying only when the page is hidden and including sub-agent notifications. Developers can install it via npm and must grant notification permissions in their browser.
Read Moredsh-dev-crew: A Subagent Plugin for Role-Based Work Assignment
The `dsh-dev-crew` plugin for DeepSeek Harness uses a multi-role subagent mechanism to dispatch different models by role within a single session to complete complex development tasks. It includes built-in roles and methodologies such as implementer and reviewer, supports Web and YAML configuration, and automatically manages artifact directories. Subagents return results via settlement notifications or the Report tool. During installation, note host version limitations, file modification permissions, and date accuracy.
Read Moredsh-webroad: Free Remote Access to DSH Web (Tailscale Funnel)
dsh-webroad is a DeepSeek Harness plugin that leverages Tailscale Funnel to provide a free and simplified web-based remote access solution. It supports QR code login over LAN and public networks, and includes secure and stable features such as PIN authentication, transparent WebSocket forwarding, automatic reconnection, and traffic compression. The `-w` parameter is required during installation. Once installed, you can manage its status via the settings page entry or the CLI. It is ideal for home and general users who want to avoid domain name configuration.
Read Moredsh-remote-access: Public Internet HTTPS Remote Access for DSH Web
The dsh-remote-access plugin is designed to solve the issue that the DeepSeek Harness web interface cannot be accessed from outside in an intranet environment. Using technologies such as Cloudflare tunnels, a password proxy, and mobile compatibility patches, it provides DSH with public HTTPS remote access. The plugin runs with the permissions of the current process, making it suitable for advanced users who need a stable public endpoint. Consult the plugin directory or GitHub repository for installation instructions, and be sure to review the source code and security before use.
Read Moredsh-skill-cockpit: Comprehensive Management Panel for DSH Agent Skills
`dsh-skill-cockpit` is a web plugin for DeepSeek Harness designed to address issues such as scattered skills, difficulty in remembering them, and “zombie skills.” It provides a centralized dashboard with full-view inspection, Chinese notes, search management, and usage statistics. Additionally, the plugin includes local testing capabilities that use vocabulary scoring to help validate skill trigger logic, enabling developers to organize their skill library more efficiently.
Read MoreIntroducing the DeepSeek Harness Plugin: dsh-rolehub-bridge
`dsh-rolehub-bridge` is a DeepSeek Harness plugin that provides a native compatibility layer for RoleHub, supporting secure loading and validation of roles. It ensures secure role loading through Hub discovery, integrity checks, policy binding, and session management. Installation requires specific Node.js and Harness versions, and the plugin supports refreshing, inspecting, and starting role sessions via the command line or UI. Configuration options can restrict capability sets, and the plugin does not provide model-facing role installation tools. After installation, users can use native controls and Host commands to create sub-sessions while maintaining context independence.
Read Moredsh-quicksight: DeepSeek Harness Dual-Layer Image Recognition Plugin
The dsh-quicksight plugin in DeepSeek Harness (DSH) supplements image recognition for text-only models using a two-layer "local OCR + vision model" strategy. Maintained by Isanti2016 and released under the MIT license, the plugin follows a "fast first, accurate later" approach: it first attempts rapid local text extraction with RapidOCR, which takes about 2–3 seconds, runs fully offline, and incurs no cost. If the result is insufficient—such as too few extracted characters—it falls back to a vision model for deeper analysis. Installation requires configuring a Python environment to run Tier-1, and optionally configuring modlens to run Tier-2. The plugin does not listen on any ports, and Tier-1 processes data purely locally to protect privacy. It supports common image formats but does not support PDFs.
Read Moredsh-opencodeproxy: DSH Plugin: Manage the Local opencode-free-proxy Service in the Settings Page
dsh-opencodeproxy is a DSH plugin developed by infexpense for managing a local opencode-free-proxy service. The plugin controls service start/stop via the settings page, supports starting with DSH, and enables automatic or manual refreshing of the model list. It provides core features such as connectivity testing and taking over existing processes, along with configuration file management. During installation, pay attention to the dependency link for `@deepseek-ai/schemastery`, and restart the service after making code changes.
Read Moredsh-tool-accurate-vision: DeepSeek Harness Image Spatial Reasoning Tool
`dsh-tool-accurate-vision` is a plugin for DeepSeek Harness (DSH) designed to address the pain point that general-purpose vision models only provide descriptions lacking geometric facts. The tool processes images through an OpenAI-compatible interface and returns structured notes, normalized bounding boxes (0-1000), and OCR information, encapsulating the data as `<vision-context>` XML for text models to parse. Its core capabilities include structured output, SVG visualization feedback, and built-in edge distance calculation functions. By converting visual impressions into computable geometric data, the tool enables models to perform rigorous precise calculations (such as node distances), significantly improving the verifiability and accuracy of the results.
Read Moredsh-deepsea: Turn Session Context into Deep-Sea Slacking
`dsh-deepsea` is a web plugin for DeepSeek Harness that transforms dialogue context pressure into a deep-sea fishing game. The hook depth changes in real time according to context length. The game features four water layers, five oceans, and independent creature pools and holographic card generation. Cards are generated by the MiniMax model and their textures are baked using Python. Caught fish can be released into a fish pond world. The plugin requires setting up a MiniMax API key and a Python environment, supports cloud synchronization of game results, and is designed to intuitively demonstrate AI processing difficulty through gamification.
Read Moredsh-ui-zoom: Browser-Style Zoom for DeepSeek Harness and DSH Desktop
`dsh-ui-zoom` is a plugin that provides browser-style UI zooming for DeepSeek Harness and DSH Desktop. It supports zoom adjustments from 50% to 200% via keyboard shortcuts (Ctrl +/-, mouse wheel) or the bottom-right widget, with settings persistently saved by source. Implemented using CSS zoom, the plugin does not affect fixed panel rendering and supports Chinese, English, and Russian. An installation command is included, aiming to improve interface readability at high resolutions.
Read Moredsh-model-picker-search: DeepSeek Harness Model Picker Enhancement
`dsh-model-picker-search` is a plugin developed for DeepSeek Harness and DSH Desktop to address the low efficiency of native menu browsing in multi-model scenarios. It is fully client-side, supports searching and filtering by name, ID, description, and provider, and includes multi-language support and result counting. By using a registration mechanism to override the native menu, it also provides fault-tolerant fallback options. Users can install it separately for the Web or Desktop Profile, filter models by clicking chips, and clear conditions by pressing `Esc`, improving efficiency when working with hundreds of models.
Read More