dsh-paper-design: Paper Design MCP Bridge Plugin
`dsh-paper-design` is a plugin for DeepSeek Harness (DSH) that integrates the design capabilities of Paper Desktop into DSH through the MCP protocol. Installation requires modifying `package.json` and restarting the process. The plugin mainly provides native tool mapping, image screenshot forwarding, Cursor-style design guidance, and three core skills. It also supports OAuth authentication management and offers commands for reconnection and authentication status checks. Before use, ensure that Paper Desktop is running and that the DSH configuration is correct.
Read Moredsh-reasoning-effort
The DSH Web interface has limited native support for reasoning effort, and the `dsh-reasoning-effort` plugin provides fine-grained control for this. The plugin defines reasoning levels (such as `low`, `medium`, and `high`) on the settings page and embeds a slider in the model selector, allowing users to adjust reasoning effort in real time. It only displays configured models and supports mouse, keyboard, and scroll wheel interactions. The Web service must be restarted after installation. Before use, ensure the model is configured under the `llm-pi-ai` provider and that the adapter is compatible with the selected level.
Read More@trapstreet/dsh-trapstreet: Check Plugin Status and Public Evaluation Board
`@trapstreet/dsh-trapstreet` is a helper plugin for DeepSeek Harness, designed to address plugin load-status verification and public evaluation board queries. It reads Harness’s internal state to provide local plugin health checks (including category identification) and remote evaluation board queries. Installation requires attention to dependency resolution issues. It supports log configuration via environment variables and performs read-only operations only, without account login or local file writes.
Read MoreDeepSeek Harness (DSH) Web Plugin: Prompt Panel Collapse Plugin
This plugin adds a collapse button to the prompt panel header, allowing the panel to shrink into a compact bar and free up visible space for the conversation history. Clicking the icon collapses or expands the panel, and the collapsed state is automatically saved. The plugin supports both Chinese and English, is implemented entirely on the frontend, does not rely on backend services, and is compatible with the DeepSeek Harness Web environment.
Read Moredsh-model-vendor-badge: Displaying the Current Model Vendor in DeepSeek Harness
`dsh-model-vendor-badge` is a DeepSeek Harness plugin developed by Townrain. It displays the provider behind the current model to the left of the Web compiler's model selector. The plugin is registered under `conversation.input.right`, reads the same `ModelDirectory` data source as the session, and updates the vendor pill display in real time. Known providers are shown with a brand-colored dot, while unknown ones use a neutral dot. Installation uses the `dsh plugin` command, and building depends on a local DeepSeek Harness repository. The plugin runs entirely on the client and leaves no residue after uninstallation.
Read Moredsh-fetch-models-unselect-all: One-Click Clear Model Selection List
This plugin is designed for DeepSeek Harness to address the cumbersome process of bulk-resetting selections in the “Fetch Available Models” dialog. It adds a “Deselect All” button to the action bar at the bottom of the dialog, allowing users to clear all selected models with a single click. It uses DOM listeners and React state synchronization to ensure that actions remain consistent with the UI state. The tool runs entirely on the client side with no side effects; users can install it via a command and use it after restarting the service. It is well suited for scenarios requiring frequent resets of the model list.
Read Moreharness-relay-mcp: Delegating and Monitoring the DeepSeek Harness via MCP Agents
harness-relay-mcp is an open-source project designed to integrate DeepSeek Harness's workflow execution capabilities into MCP agents such as Codex, Claude Code, and Cursor. As a control plane, it does not run model inference; instead, it handles session management, permission control, and tool dispatch. The plugin supports native session persistence, asynchronous lifecycle management, tiered permissions, and multimodal input, interacts with external agents via the MCP protocol, and addresses the challenges of long-running tasks and fine-grained permission management.
Read Moredsh-vision-api-localorweb: Integrate a Local Vision Model API and Equip the DSH Assistant with a Pair of “Local Eyes”
DeepSeek Harness (DSH) does not support image recognition by default. This plugin enables access to local models to read image content while preserving privacy. It supports analyzing images through local OpenAI-compatible endpoints, such as LM Studio or Ollama, keeping all data on the local machine. Users can configure the endpoint address, model name, and API key in the settings interface without modifying configuration files. After installation, simply provide an image path in the conversation to obtain a text description. Note that direct image sending may be limited; using the file path is recommended, and the local service must be compatible with the standard interface.
Read Moretipoli5890/dsh-file-mention: Introducing the DSH Web Plugin
The dsh-file-mention plugin provides secure file mention and upload capabilities for the DeepSeek Harness Web composer. Triggered by `@`, it lets users quickly index and search workspace files, with Git-aware path reference management. It also supports drag-and-drop or paste uploading for non-image files, and includes features such as upload progress, quota limits, and a mention tray. Installation is simple, making it ideal for developers who frequently reference local code or documentation.
Read Moredsh-client-ui-trace-graph: DSH Web Trace Graph Visualization
The DSH Web GUI plugin adds an interactive ReactFlow DAG alongside the official “Trajectory” tab, using official data to visualize execution history. Nodes represent trajectory records (such as SYSTEM, USER, and TOOL), while edges indicate temporal order (gray dashed lines) and structural relationships (colored solid lines). Users can click a node to view JSON details and switch between dagre and d3-force layouts. Installation requires Node.js ^22.19 or >=24, with attention to pnpm version compatibility issues.
Read Moredsh-llm-local-token: Using Local CLI OAuth Tokens for DSH Model Routing
The `dsh-llm-local-token` plugin resolves the issue where DeepSeek Harness (DSH) cannot use local Codex or Claude OAuth tokens. It reuses local subscriptions as DSH's model route (such as `openai-codex`) and displays real-time usage in the Composer toolbar. The plugin supports automatic token refresh and automatic configuration injection, as well as GLM quota display. Note that GLM quota display depends on the `sse` protocol and requires Node.js >= 22.13.0.
Read MoreA-Share Stock Research Assistant - DeepSeek Harness (DSH) Plugin: Stock Search / Historical Announcements / Financial Reports / Company Profiles / Signal Explainers
The `dsh-astock-research` plugin for DeepSeek Harness (DSH) is designed to fill the gap in A-share data querying. It provides core capabilities including individual stock search, stock profiles, plain-language interpretation of announcements, and interpretation of financial statement figures. Installation requires meeting the Node.js version requirements and configuring an API Key in the Web UI. The plugin covers data for approximately 5,000 A-share stocks and supports querying arbitrary historical periods. For complex computations such as strategy backtesting, the tool guides users to the AIHEY Trading Simulation Assistant. The data is provided for research reference only and does not constitute investment advice.
Read MoreDeepSeek Harness Plugin: dsh-computer-use
`dsh-computer-use` is a native Windows desktop automation plugin for DeepSeek Harness (DSH), designed to provide a lightweight control solution. It directly calls the DSH Cordis tool registry and relies on `user32.dll` to perform operations such as taking screenshots, clicking, dragging, and typing, without relying on the MCP protocol, with the native approval service enabled by default to ensure security. The plugin supports sending screenshots to external vision models for analysis. Installation requires specifying the DSH configuration file and restarting; it is currently supported only on Windows, and it is recommended to use `/permission workspace-write` in conversations to prevent operations from being denied.
Read Moredsh-coyote: DG-LAB Electrical Stimulation Plugin for DeepSeek Harness
`dsh-coyote` is a plugin for DeepSeek Harness (DSH) that connects to DG-LAB electrical stimulation devices, using a plugin-based architecture. It provides a unified safety boundary for Agent tool calls and web GUIs, including eight model-side tools, a visual control panel, and automatic stimulation functionality. The plugin enforces strict soft limits, hard limits, rate limits, and session cooldowns, includes 12 built-in waveform presets, and can be paired with the DG-LAB App via QR code. Usage requires Node.js >= 22 and involves physical risks, with operations strictly constrained by safety mechanisms.
Read Moredsh-message-annotator: Adding Citation and Annotation Cards to DeepSeek Harness
`dsh-message-annotator` is a client plugin for DeepSeek Harness designed to enable precise annotations of Agent responses. It supports citing and annotating selected text or full messages, while generated annotation cards can be temporarily stored above the input box. Users can choose to submit annotations together with a draft to the Agent or simply insert the text for review. The plugin offers a bilingual Chinese-English interface, and its data is only valid within the current session.
Read Moredeep-structural-analysis-skill: Interdisciplinary Deep Structural Analysis Skill
`deep-structural-analysis-skill` is an interdisciplinary deep structural analysis tool designed to improve analytical depth and reliability through a standardized process. The skill includes the “Attack Loop Protocol” and the “Four-Check Deduction,” requiring the exposure of default positions and adversarial challenges while using 16 disciplinary lenses and 10 core tools (such as three-way problem analysis and the 80-20 rule) to conduct multidimensional analysis. It emphasizes fact binding and confidence labeling, and is suitable for exploring the attribution of complex systemic issues in social, economic, and other domains, but not for simple factual queries.
Read Moredsh-rules-paths: Adding Claude Code-Style Path Rule Injection to DeepSeek Harness
The **dsh-rules-paths** plugin is a core extension component of DeepSeek Harness (DSH), designed to address the problem of models having difficulty remembering specific specifications in large projects. It is based on Claude Code's `paths:` mechanism; when the model reads files that match a rule, it injects the rule body into the context. Core features include: defining `paths` glob matching patterns via YAML Frontmatter, supporting both global and project-level rule scanning (`.dsh/rules` and `.claude/rules`); and providing SHA-1 deduplication, byte budget control, and a replacement mechanism. Installation requires using a GitHub repository command and manually activating it in a preset. The plugin requires DSH 0.1.0-rc.x (We
Read Moredsh-narrative-voice: DeepSeek Harness Request-Tool Pronoun Consistency Plugin
`dsh-narrative-voice` is a DeepSeek Harness (DSH) plugin designed to address pronoun inconsistency caused by perspective drift in the `ask_user_question` tool. During prompt construction, it rewrites tool descriptions to enforce a consistent referent for “I” and “you,” ensuring clear interaction. It supports two modes—“AI narrative” and “user narrative”—and can be switched in real time via the `/voice` command. The plugin has no dependencies; after installation, it takes effect after restarting the process. It is mainly suitable for application scenarios that require high consistency in interaction perspective.
Read MoreDeepSeek Harness Plugin: dsh-replay-lab
dsh-replay-lab is a plugin for DeepSeek Harness designed to address the difficulty of reproducing discrepancies that arise from interactions between models and the Harness. By freezing turns and isolating candidate sessions, it provides request-surface replay, A/B experiments, and multi-dimensional comparison features, helping developers analyze reasoning paths, costs, and debug long trajectories or regression issues. The plugin supports custom prompts, ensures data safety without affecting source sessions, and is suitable for in-depth analysis of the combined effects of “model × Harness.”
Read Moredsh-egress-guard: DeepSeek Harness Runtime Security Gateway
dsh-egress-guard is a runtime security plugin for DeepSeek Harness that enforces egress control within the tool invocation pipeline. It provides three core capabilities: intercepting unauthorized network targets via a whitelist before execution, automatically redacting sensitive information after execution, and recording comprehensive audit logs. The plugin runs in `monitor` mode by default and supports progressively building security policies from log analysis. This approach enhances agent security but cannot defend against process-level attacks or malicious traffic bypass protection via Base64 encoding.
Read Moredsh-fetch-third-party: Safe Third-Party Web Fetch for DeepSeek Harness
The `dsh-fetch-third-party` plugin addresses SSRF security risks in web scraping with DeepSeek Harness (DSH) through a “courier mode.” It delegates requests to user-configured third-party services instead of making direct local connections, thereby eliminating the attack surface. The plugin supports multiple providers, including Tavily and Jina Reader, as well as a custom Crawl4AI stack, and features secure credential storage, session-level budget limits, scraping caching, and structured output, ensuring safe and efficient retrieval of web content in DSH environments.
Read Moredsh-updater: Incremental Update and Rollback Plugin for DeepSeek Harness (DSH)
dsh-updater is a community-maintained incremental update plugin for DeepSeek Harness (DSH), designed to address out-of-memory (OOM) issues commonly caused by full installations. It enables fast updates by downloading only changed packages, supports one-click operation, preserves third-party dependencies, and provides automatic backup and rollback. Installation requires the use of an absolute path and enabling the plugin in the settings panel. It is suitable for resource-constrained environments such as Docker and nvm, helping ensure a secure and stable update process.
Read MoreIntroduction to the DeepSeek Harness CodeBuddy Plugin
The DeepSeek Harness (DSH) plugin “taikaikaikai-pixel/dsh-codebuddy-plugin” aims to integrate Tencent Cloud CodeBuddy capabilities into DSH, providing an unofficial but fully featured model inference and tool integration solution. It supports 18 mainstream models and includes image input, adjustable thinking intensity, and automatic routing. Key features include integration of CodeBuddy web search/scraping and image generation tools, with unified handling of OAuth login and multi-key management through a local streaming bridge. Users can configure credentials in the Web UI settings card and must run the streaming bridge locally to enable the primary chat function.
Read MoreDeepSeek Harness Universal UI Personalization Plugin (dsh-customization-settings)
The dsh-customization-settings plugin, developed by T-MKT, aims to provide a generic entry point for UI personalization settings in DeepSeek Harness (DSH). The plugin currently has a basic framework in place, adds an "Appearance" category to the settings panel, and integrates related UI components, with plans to support features such as wallpapers and theme colors later. As it has not been published to npm, it must be installed from a local repository and added to the `bundles` list in the Web Profile. Development depends on DSH client components and supports `pnpm` builds. The plugin is still in development and has limited maintenance by the author, so its source code and license should be checked before installation.
Read Moredsh-session-plus: Session Archiving and Branching Plugin for DeepSeek Harness
`dsh-session-plus` is a DeepSeek Harness plugin designed to complement the limitations of the official archiving feature. It provides session archiving, restoration, deletion, and conversation branching through a sidebar directory. The plugin supports idle waiting to prevent log corruption during archiving and disables archiving while tools are being invoked. It does not modify core files and relies on public APIs, ensuring compatibility and security.
Read More