dsh-browser-annotate: A Browser Annotation Tool Integrated with Real Chromium
`dsh-browser-annotate` is a browser interaction plugin for DeepSeek Harness (DSH), designed to address the lack of intuitive feedback and advanced debugging capabilities in agents. It provides a real Chromium environment and CDP access through an Electron desktop app, supporting 12 operational tools, including navigation, screenshots, clicking, dragging, popup handling, and custom code execution. A key feature is its right-click annotation mechanism, which lets users enqueue element information—such as selectors, screenshots, and styles—and inject it into the conversation, with a dual-pane layout and local communication support. Installation requires DSH plugin commands. Known limitations include the inability to close tabs automatically and the lack of support for creating new browser contexts.
Read MoreTaishan Vision: Enabling Pure-Text DeepSeek Harness Models to Understand Images
Taishan Image Recognition is a free plugin designed specifically for DeepSeek Harness (DSH). By invoking Zhipu GLM’s vision model, it injects image understanding capabilities into text-only models without replacing the main model. It supports one-click enable/disable, secure API key configuration, and card-panel management, and offers Chinese text recognition with automatic fallback. Users only need to send an image in a conversation, and the plugin automatically injects instructions for the model to describe it. This makes it suitable for scenarios where text-only models in local private deployment environments need additional image understanding capabilities.
Read Moredsh-chat-cost: DeepSeek Harness Chat Cost Tracking Plugin
`dsh-chat-cost` is a plugin for the DeepSeek Harness (DSH) Web UI, designed to address the lack of transparency in cost tracking during agent development. It provides real-time cost breakdowns, multi-provider pricing support, local JSONL cost logging, and budget planning tools. After installation, configure log writing and language options, and refresh the price catalog with `npm run prices`. The plugin displays costs by conversation tree hierarchy, but note potential discrepancies with platform billing and data loss that may result from session interruptions.
Read MoreDeepSeek Harness Task Notification Plugin: @idoall/dsh-notify
`@idoall/dsh-notify` is a plugin for DeepSeek Harness designed to address the lack of intuitive task-status feedback in long conversations. It offers enhanced and soft notification styles, a collapsible stack, direct navigation, and sound feedback. Notifications are triggered only when the Agent enters an idle state, and users can respond directly within the card. The plugin registers only a page overlay, stores data in memory, and requires DSH version `>=0.1.7-rc.1` and `<0.3.0`.
Read Moredsh-TUI Secret Guard: Intercepting High-Confidence Secrets in dsh-TUI
dsh-TUI Secret Guard is a dsh-TUI plugin designed to prevent leakage of sensitive information. It intercepts high-confidence credentials before sending a prompt, scans sensitive assignments such as API keys, and never stores content. Installation requires declaring specific permissions and supports YAML configuration. The plugin performs local scanning, intercepts decision events, and includes a compatibility fallback mechanism. Blocked prompts must be edited manually to avoid misunderstandings caused by silent changes.
Read MoreDeepSeek Harness Plugin: AI Aggregation Gateway Visualization Console
The DeepSeek Harness plugin provides multi-channel model reverse proxying and a visual control hub, integrating environment deployment, OAuth authorization, intelligent multi-account rotation pools, quota monitoring, and intelligent proxy optimization. Supporting multiple channels such as Google and OpenAI, it exposes OpenAI-compatible interfaces for DSH integration and features high-precision quota dashboards, ultra-fast SWR responses, and failover capabilities. After installation, accounts can be onboarded with one click by authorizing them through the Web interface. It incorporates security designs such as local binding, key separation, and self-healing mechanisms, making it suitable for managing locally deployed large language models.
Read Moredsh-latex2office: DeepSeek Harness LaTeX-to-Office Equation Plugin
The DSH plugin `dsh-latex2office` is specifically designed to address the challenge of converting LaTeX formulas into Office documents. It supports either native OMML editing or high-resolution rendering. Key features include batch atomic operations, academic numbering for typesetting, automatic backup read-back verification, and isolated secure rendering. The plugin can be invoked through JSON configuration and supports creating new documents or inserting into existing ones. It has zero npm dependencies and requires Python, Pandoc, and LibreOffice to run, delivering an end-to-end solution from integration with formula editors to batch typesetting implementation.
Read MoreDeepSeek Harness Mobile Pairing Floating Button Plugin (DSH-QrPairing)
The DSH-QrPairing plugin adds a floating entry in the bottom-right corner of the DeepSeek Harness WebUI, resolving the need to frequently switch browser tabs during remote pairing. It integrates a pairing QR code viewer, allowing users to view the code without leaving the current page. The plugin depends on the `dsh-remote-link` gateway and uses a same-origin proxy to fetch data in real time. By rendering the QR code with an `<img>` tag instead of an `<iframe>` in the Electron environment, it ensures desktop compatibility and reliable loading. Make sure the relevant gateway plugin is installed before setup.
Read Moredsh-chatbi: A Fully Self-Contained DeepSeek Harness ChatBI Plugin
`dsh-chatbi` is a client plugin for DeepSeek Harness that runs ChatBI logic locally. It supports managing and invoking SQLite databases, MCP tools, and APIs, with all configuration stored in a local JSON file. The plugin provides a rich set of Agent tool interfaces, enabling agents to directly execute SQL or invoke MCP tools or APIs. Although MySQL/PostgreSQL metadata storage is supported, actual execution requires manually configuring the appropriate drivers. Developers need a Node.js ≥ 22 environment, install it using the `dsh plugin` command, and configure it in the settings page.
Read MoreDeepSeek Usage Panel: Balance and Token Dashboard (dsh-usage)
The DSH plugin `dsh-usage` solves the issue where regular plugins cannot track API calls made via the web and client by calling DeepSeek’s official interfaces. It supports viewing the full balance, token usage, and costs, and provides multi-time-range charts along with queries for models and tools. After installation, users must obtain a platform session token for configuration; the token is stored only locally and remains secure and transparent. This plugin is suitable for users who need to monitor consumption across all channels in a unified way, ensuring real and accurate data.
Read MoreDeepSeek Harness Plugin: dsh-quant-factor-pipeline (Quantitative Factor Research Pipeline)
The `dsh-quant-factor-pipeline` plugin for DeepSeek Harness (DSH) is designed to encapsulate the end-to-end quant factor research process into an orchestratable workflow tool. It provides complete functionality, from panel data retrieval, factor calculation (with IC gating), and coarse-grained Regime detection, through backtesting, to project state persistence. Installation requires manually running the relevant npm command, and the plugin runs on Node 22.22.2 with zero dependencies. It uses environment variables to specify data directories and log paths, and strictly separates valid data endpoints from restricted ones to ensure compliant data fetching. The author advises version locking and clarifying the responsibility boundary with `dsh-quant-data-mcp`.
Read Moredsh-industry-graph-mcp: Zero-Dependency A-Share Industry Chain Knowledge Graph
`dsh-industry-graph-mcp` is a Node.js-based A-share knowledge graph MCP server. It provides 7 local tools for querying individual stock industries, industrial chain positions, peer competitors, and concept constituents. With zero dependencies and no API required, the plugin uses a local graph service to provide Agents with factual grounding for quantitative investment research, supporting seamless integration into DeepSeek Harness.
Read Moredsh-hr-payroll-mcp: Localized Payroll Calculation MCP Service
This is a local-first, general-purpose HR payroll MCP service designed to address the complexity of enterprise payroll rules and the security concerns of uploading data to the cloud. The plugin runs in zero-dependency Node.js ESM mode and provides generic header adaptation, enterprise configuration read/write, a statutory calculation engine, secure formula evaluation, and payslip export. It supports social insurance and housing fund parameters for 12 cities (2026-ref version) and allows users to extend coverage to cities not yet included via the `refresh_statutory` tool. During deployment, configure an absolute path or environment variables to ensure payroll logic runs locally, meeting offline audit and data security requirements.
Read Moredsh-docx-mcp: A Dependency-Free MCP stdio Service for Generating .docx Files
`dsh-docx-mcp` is a local, zero-dependency plugin for DeepSeek Harness that generates real `.docx` files. It supports headings, paragraphs, lists, and tables, and runs without installation. Agents can invoke it via the `docx_generate` tool by providing title, content, and structure parameters to directly output document files. Currently, the plugin requires manual path configuration, making it suitable for quickly generating contracts, reports, and similar documents.
Read Moredsh-dcs-engine: Dynamic Capability Synthesis Engine
`dsh-dcs-engine` is the Host composition layer of the DSH protocol stack, designed to address the challenge of manually orchestrating multiple basic tools when building complex business plugins. The engine can synthesize discrete basic tools into reusable DSH Capabilities at runtime and supports generating complete plugin skeletons through a Materializer. It uses CDP for semantic scoring, reuses the intent network for topology planning, writes decisions to CSP, and finally enforces artifact security review via ASP. Plugin installation is simple and supports synthesis strategies such as rule/llm, making it the core demonstration vehicle for validating the collaborative capabilities of DSH's five-piece set.
Read MoreDeepSeek Harness Context-Aware Web Search Plugin
`dsh-context-aware-search` is a plugin for DeepSeek Harness (DSH) designed to address the issue of irrelevant general search results. It rewrites queries by reading conversation context, uses TF-IDF to re-rank results and mark their credibility, and supports one-click source summarization. The plugin uses DuckDuckGo by default and requires no API key, significantly improving search precision in specific technical scenarios, such as programming documentation retrieval.
Read Moredsh-cn-fixedincome-mcp: Local-First Analysis Engine for Convertible Bonds/Bonds
helibeiqi/dsh-cn-fixedincome-mcp is a zero-dependency Node ESM MCP plugin for DeepSeek Harness (DSH), focused on local computation of fixed-income data. It does not rely on external APIs, and its core capability is to quickly derive metrics from locally provided terms. The plugin provides valuation and strategy indicators such as convertible premium, YTM, and duration, as well as risk assessments for events such as downward revision dynamics and forced redemption. Through tools such as `cb_analytics` and `cb_screen`, it supports batch screening and in-depth analysis of local seed data. This plugin implements an offline, key-free quantitative investment research solution and is open-sourced under the MIT license.
Read Moredsh-cn-disclosure: Zero-Dependency Local-First Structured Extraction of A-Share Announcements and Annual Reports
`dsh-cn-disclosure` is a DeepSeek Harness plugin dedicated to localized structured extraction of A-share disclosures. Unlike tools that rely on external APIs, it runs entirely offline, supports 17 disclosure event types and 9 extraction fields, and outputs JSON data ready for event studies. Users can extend it by editing configuration files or rule sets, making it suitable for large-scale text analysis scenarios that require privacy protection and local processing.
Read MoreHDNRAY/dsh-plugin-colorguess: Play a Quick Color-Mixing Game While Waiting for Model Output
The `dsh-plugin-colorguess`, developed by HDNRAY, is an entertainment plugin designed for DeepSeek Harness. It renders a button in the session header and uses an iframe to embed the “Color Mix: Three Choices, One Pick” game hosted on GitHub Pages, providing instant fun. The plugin supports a draggable floating window, light/dark theme synchronization, and trace-free interaction, leaving no residue after it is closed. Installation requires the `web` profile and a restart. It runs only in the browser with an internet connection, requests no special permissions, and is privacy-safe.
Read MoreIntroduction to the DeepSeek Harness SSH Tool Plugin (dsh-tool-ssh)
`dsh-tool-ssh` is an SSH tool plugin for DeepSeek Harness (dsh), designed to address remote host operation requirements. By invoking the local OpenSSH client, it achieves **zero agent deployment**, meaning the target machine only needs to support SSH login. The plugin provides three core functions: executing commands (`ssh_run`), uploading (`ssh_upload`), and downloading (`ssh_download`). During installation, ensure that both dsh and OpenSSH are installed locally. When using it, note that batch mode is enabled by default to avoid interactive prompts; key-based authentication is recommended, and the output stream has size limits to prevent memory overflow.
Read Moreoh-my-dsh: DeepSeek Harness Multi-Agent Orchestration Layer
`oh-my-dsh` is a multi-agent orchestration plugin based on DeepSeek Harness (DSH), designed to replicate proven workflows. By encapsulating 30 auto-loaded skills and 13 slash commands, it enables zero-configuration team-first orchestration, automatic parallelization, and natural language interaction. Key features include inline HUD progress visualization and experience-driven learning. Built on DSH primitives with no additional runtime overhead, it is well suited for rapidly building multi-agent collaboration systems.
Read Moredsh-fight-scene-director: DSH Native Fight Scene Director Skill
The `dsh-fight-scene-director` is a native plugin for DeepSeek Harness (DSH), designed to port the Codex skill for choreographing high-quality AI video fight scenes. Installed via the plugin manager, it supports designing fight sequences from text briefs, generating trajectory maps, reading assets, and adapting to multiple combat types. It outputs structured Chinese prompts compatible with Seedance 2.0 and similar models, ensuring action causality and continuity while providing pacing control. Caveats include single-Profile limitations, local installation risks, and configuration validation requirements.
Read Moredsh-task-orchestrator: Automated Plan-First Task Orchestration
`dsh-task-orchestrator` is a plugin designed for DeepSeek Harness (DSH) to automate task decomposition and dependency management in complex workflows. It adopts a plan-first mechanism that allows the parent model to create and assign TODO plans, then processes long-chain tasks efficiently using dependency-aware queues and isolated execution environments. The plugin supports single-agent and multi-agent parallelism and provides state snapshot functionality for workflow recovery. After installation and enablement, it is well suited to complex scenarios involving coordinated subtasks with dependencies, significantly improving processing efficiency and accuracy.
Read Moredsh-session-resilience: DSH Safe Restart and Policy-Based Session Recovery
`dsh-session-resilience` is a configuration plugin for DeepSeek Harness (DSH), designed to address session interruptions caused by host restarts, network disruptions, or token limits. Through safe restarts, token-aware reconnection, and policy-based session recovery, it ensures that DSH can seamlessly resume tasks after interruptions. Its core features include local control, four preset recovery strategies (Safe, Balanced, Long task, and Manual), and native tool-calling support. After installation, users can manage session recovery uniformly via the settings interface; the system reports readiness only after successful connection verification, significantly improving the stability of agent task execution.
Read Moredsh-wemedia-workbench: DeepSeek Harness — A Local-First Self-Media Workbench
`dsh-wemedia-workbench` is a local-first self-media workbench plugin for DeepSeek Harness. It aims to address common issues in the native interface by providing core capabilities such as content organization, version control, review evidence, and multi-channel publishing. The tool supports local content libraries, immutable revisions, batch WeChat draft submission, and adapters such as WeChat Official Accounts, emphasizing workflow security and controllability. The project is in an early development stage, requiring users to configure accounts and permissions on their own, and it is not an official DeepSeek product.
Read More