sensedeal/cue-skills: Cue Agent Skills and DSH Plugin Pack
sensedeal/cue-skills is an open-source agent skill package designed to inject domain-specific capabilities into agents such as DeepSeek Harness and Claude Code. It includes instruction packs and a DSH plugin that can be loaded without modifying agent code. Core features cover writing Cue research templates, parsing HTTP/media files, querying public data (finance/regulatory), and supporting Playbook scenarios. Users can quickly install skills via the CLI, while noting API key configuration and being aware that the DSH root package lacks SSRF protection and may introduce permission risks.
Read Moredsh-acl-sandbox-patch: DSH 0.1.7 Windows Sandbox Non-System Drive Workspace Fix
dsh-acl-sandbox-patch is a temporary workaround plugin for DSH 0.1.7 that addresses the complete failure of sandboxed commands in workspaces located on non-system drives. By wrapping the underlying API, it resolves the issue where `SetNamedSecurityInfoW` returns Win32 error 5 due to insufficient permissions, and handles token integrity control. After installation, DSH must be restarted to restore normal command execution; however, this compromises part of the integrity-labeling mechanism and carries risks of slower initial loading and abnormal behavior in older sessions.
Read MoreDeepSeek Harness Plugin: dsh-plugin-manager-companion
`dsh-plugin-manager-companion` is the official companion to the DeepSeek Harness (DSH) plugin manager. Through its "everything is a plugin" architecture, it fills gaps in the original manager's pre-installation quality gates, in-depth environment diagnostics, plugin marketplace, and upgrade workflows. Core features include pre-installation quality gating and five-layer environment diagnostics (dependencies, composition, runtime, etc.); an environment console for profile management; and community plugin marketplace and skill preset management. The tool also supports secure upgrades and command-line operations, aiming to reduce environmental failure rates, standardize plugin acquisition workflows, and suit developers who require rigorous environment management.
Read MoreA-Share Workbench: Turning the DeepSeek Harness Session Window into a Market-Watching Execution Desk
`@lisonevf/dsh-stock-panel` is a web plugin for DeepSeek Harness (DSH), designed to transform the AI chat window into a schedule-organized A-share market monitoring and execution console. It provides features such as a market dashboard, post-market reviews, sector mining, individual stock workspaces, and AI-assisted analysis. The interface strictly controls information density and follows a read-only results design principle, helping users address key questions and make decisions during trading hours. The plugin relies on user-configured data sources, does not include built-in market data, and is for analytical reference only, with no promise of returns.
Read Moredsh-memory-layer: A Local Memory Layer for DeepSeek Harness
`dsh-memory-layer` is a local memory plugin for DeepSeek Harness, designed to solve cross-session memory retention. It is based on single-file SQLite storage, requires no cloud dependency or vector retrieval, and runs fully locally with zero service dependencies. It performs deterministic recall using trigger signals such as file names and tool names, supports memory management through the `/memory` command, and includes cleanup mechanisms such as automatic deprioritization, tombstone deletion, and workflow chain promotion. It is suitable for developers who prioritize high privacy, low latency, and explainability.
Read MoreDeepSeek Harness Reasoning Level Auto-Selection Plugin
The `dsh-plugin-jev-effort-selector` plugin uses the Jev System One model to automatically assess message complexity, enabling intelligent derivation of DeepSeek Harness reasoning levels. It reads model-declared levels, uses the context envelope to determine task continuity, and ensures unfinished tasks are not set below the previous tier. With session isolation, manual-selection priority, and silent fallback on failure, it addresses the inefficiency and forgetfulness of manual switching, making it suitable for DSH users who need fine-grained control over reasoning costs.
Read MoreDeepSeek Harness Plugin: dsh-cyberwhale
The `dsh-cyberwhale` plugin for DeepSeek Harness aims to address the style mismatch between the software interface and the local operating system. Through a dual-platform localization design, it provides macOS- and Windows-native settings interfaces, 27 theme color palettes, font customization, and background settings, with support for picking colors from wallpapers. In addition, the plugin includes an interactive blue whale desktop pet that can follow the cursor, show speech-bubble tips, and more. Maintained by CyberWei922 under the MIT License, it is easy to install and supports local development and debugging.
Read Moredsh-harness-jarvis: Desktop Floating Assistant and Particle Orb
Jarvis is a desktop floating-window assistant and pet based on DSH Desktop. It takes the form of a metallic orb and offers features such as invoking conversations via global hotkeys, desktop permission approval, and voice greetings, aiming to reduce context-switching costs. The plugin requires macOS 14 or later. Data is stored locally only, with no cloud synchronization or sensitive permissions involved, aiming to optimize the agent debugging experience through a persistent interaction layer.
Read Moredsh-context-imports: Inject Files Referenced by AGENTS.md at Session Startup
The `dsh-context-imports` plugin addresses the issue where models in DeepSeek Harness cannot read `@path` references inside `AGENTS.md`. It automatically scans and injects the content of referenced files into the session context, supporting recursive expansion, byte-budget control, and multilingual configuration. The plugin is automatically triggered in scenarios such as session startup; by resolving relative paths and skipping code blocks, it wraps file content in a `<system-reminder>` message and requires no additional configuration to use.
Read Moredsh-auto-retry: DSH Workflow Auto-Recovery Plugin
The dsh-auto-retry plugin is designed to address network instability, server-side errors, and interrupted streaming output in DeepSeek Harness workflows. It provides request-level supplementary retries (covering eight error types), turn-level automatic continuation with a circuit breaker mechanism, and unresponsive watchdog functionality. The plugin adopts a main/sub-agent divide-and-conquer strategy, offering real-time status display, data dashboards, and input optimization to automate cumbersome manual intervention. Users can install it via a command and flexibly configure retry counts, delays, and policies in the settings, making it suitable for unstable network environments or long-running task scenarios.
Read MoreTheOne: A Single Conversational Entry Point for DeepSeek Harness
TheOne is a memory management plugin for DeepSeek Harness (DSH), designed to resolve chaotic session management and mixed contexts. It follows a “One chat for everything” approach: users only need to interact through a single main chat window, while TheOne automatically creates independent sessions for different topics in the background and maintains context isolation. Its core features include single-entry chat, automatic routing, unified model invocation, and session directory management. It is easy to install and directly uses the API key already configured in DSH. TheOne does not delete records; it only organizes existing sessions, helping users manage multi-task conversations efficiently.
Read MoreDeepSeek Harness Plugin: weibaohui/dsh-sync
weibaohui/dsh-sync is a DeepSeek Harness plugin designed to resolve data inconsistency issues in multi-machine development. It supports Git, WebDAV, and local folders as backup protocols, enabling synchronization of skills, sessions, settings, and plugin manifests. The plugin features AI-driven intelligent alignment and conflict resolution, supports private repository validation and token isolation, and provides local-first snapshots with cloud recovery, delivering a secure and reliable solution for multi-machine collaboration.
Read Moredsh-sidebar-size: DeepSeek Harness Right Sidebar Size Persistence Plugin
The `dsh-sidebar-size` plugin aims to solve the issue in DeepSeek Harness of having to repeatedly adjust the right sidebar width when frequently switching tabs. It supports remembering sidebar sizes by tab type, features an inheritance mechanism and a unified width mode, and allows users to view status, export logs, and migrate data from Settings. Users can install it by downloading a `.tgz` file or from local source code. Developed for DSH Desktop v0.2.0-rc.x, the plugin does not modify host files; however, abnormal exits may result in data loss, and compatibility with the host application version must be maintained.
Read MoreDeepSeek Harness Mathematical Modeling Team Plugin Pack
The DeepSeek Harness plugin `dsh-math-team` provides a standardized collaboration framework for mathematical modeling competitions. The plugin includes two types of Agent presets: `model-code` for modeling programming and `paper` for paper writing, supporting multi-person collaboration in independent workspaces via GitHub/Gitee. Core features include quality gates such as M1-M2/P1-P2, particularly M2 robustness attacks, independent dual-model review, and mandatory figure review, ensuring the quality of both the model and the paper. The plugin requires DSH 0.2.0+ and is installed/configured via the `link:` method.
Read MoreDeepSeek Harness Machine Learning Phase Gate Plugin (dsh-plugin-ml-phase-gate)
The machine learning stage-gate plugin for DeepSeek Harness transforms status management from subjective descriptions into “evidence-driven” tracking by dividing the project into M0–M8 stages. It reads the project status, runs user-provided validation scripts, calculates the actual stage and any gaps, and displays audit information in the DSH interface. The plugin supports command-line tools, cross-platform validation scripts, and project onboarding, making it suitable for ML workflows that require strict process management.
Read Moredsh-tool-antigravity: A High-Performance Antigravity OAuth LLM Routing and Nano Banana 2 Image Generation Capability Pack
`dsh-tool-antigravity` is a high-performance plugin for DeepSeek Harness, designed to enhance Antigravity model routing and image generation capabilities. It addresses challenges in OAuth authentication, long-text streaming, and image editing, supporting Google OAuth 2.0 PKCE authentication, the Nano Banana 2 image engine, and fine-grained inference budget control. The plugin optimizes network performance through connection pooling and implements strict permission loopback protection along with secure storage mechanisms. It is community-driven, licensed under the MIT License, and provides developers with an integrated and secure tool-calling solution.
Read MoreDeepSeek Harness LazyGit Plugin: A Web-Based Visual Git Workbench
`dsh-lazygit-plugin` is a visual Git workbench specifically built for DeepSeek Harness Web. It addresses the fragmentation of Git operations in web environments by providing core features such as commit review, branch management, conflict resolution, and AI assistance. Developers can intuitively view diffs, stage changes, and safely execute AI-generated multi-step commands, with support for Normal, Fast-forward, or Squash merge strategies. Before installation, the old plugin must be removed and the environment must be ready.
Read Moredsh-mcp-manager-plus: DeepSeek Harness MCP Server Management Plugin
dsh-mcp-manager-plus is a DeepSeek Harness (DSH) plugin maintained by appthin, designed to reduce errors when manually editing MCP protocol configuration files. It adds an "MCP Management" page to the DSH settings interface, enabling users to view, enable/disable, edit, restart, and delete MCP servers directly in the browser. Core features include expandable list views for tools, dual-mode editing (form and JSON), automatic detection of multiple configuration formats (such as TOML and JSON), and multilingual support. Before installation, users must meet the required DSH version and environment; after installation, MCP servers can be managed efficiently across their full lifecycle through the visual interface.
Read Moredsh-chat-thinking-editor: Edit AI Responses and Chain of Thought Directly in DeepSeek Harness
`dsh-chat-thinking-editor` is an official plugin for DeepSeek Harness, designed to address model "stubbornness." It allows developers to visually edit the response body and underlying chain of thought of any AI message directly in the chat interface, ensuring that subsequent generation is based on the modified content and enabling precise control. The plugin supports features such as full-text editing, history message management, null-value handling, and preservation of non-text content. Installation requires DSH version 0.1.5-rc.2 or later and can be installed via npm. It is particularly suitable for developers who need to adjust output formats in real time or correct reasoning direction during debugging.
Read Moredsh-chat-manager: DeepSeek Harness Chat Management Plugin
`dsh-chat-manager` is a user-level plugin developed for DeepSeek Harness (DSH), designed to address the cumbersome lookup and management of chat history as conversation volume increases. Through features such as a dual-pane sidebar, automatic archiving, session deletion, and session restoration, it provides intuitive session lifecycle management. It supports both the Web UI and desktop clients and aligns with DSH’s plugin ecosystem approach. Installation requires DSH version >= 0.2.0-rc.2; users must build the plugin themselves and add it to the plugin directory. In addition, active session deletion and search on Linux/macOS have caveats related to their dependency on a specific index configuration.
Read Moredsh-djy-xttsc: Global System Prompt Injection Plugin
The dsh-djy-xttsc plugin provides a visual “Big Fat Fish Instructions” configuration in the settings page to inject a global system prompt into all model requests. It supports hot reloading, persistent storage, and version compatibility, ensuring the instruction takes effect across sessions, workflows, and subagents. Installation requires command-line operations, while usage only involves enabling the option in the settings panel or modifying the instruction text. Note that the plugin content is injected as-is and conflicts with other injection sources in the same environment should be avoided. If the configuration does not take effect, enable diagnostic mode to troubleshoot.
Read Moredsh-novel-craft: DeepSeek Harness Novel Writing Plugin
dsh-novel-craft is an open-source plugin for DeepSeek Harness designed to address style transfer in AI novel writing. It captures users' reading preferences so the model can distill and reuse writing patterns, without relying on prompt engineering or fine-tuning. The plugin supports preference calibration, candidate draft generation, annotation-based refinement, plot health checks, and local pattern storage. Users can create content in the interface by marking paragraphs, extracting patterns, starting new chapters, selecting passages through card draws, and refining text according to annotations, while the main text does not enter the model context, keeping the process lightweight. The plugin depends on the DSH environment and provides a Chinese interface.
Read Moredsh-jev-plugin: A Typed Decision Plugin for DeepSeek Harness
dsh-jev-plugin is a DeepSeek Harness (DSH) plugin designed to encapsulate the structured judgment capabilities of TypeSafe Jev as tools that assist agents in making typed decisions. Its core features include exposing the three primitive tools `noul`, `choice`, and `score`, along with configurable guardrails and review mechanisms. Installation requires Node.js >= 22 and supports Web UI result display. Configuration includes API key setup and toggle controls for risk interception, result correction, and offline rules, making it suitable for DSH projects that require strict risk control and structured output.
Read Moredsh-workbuddy-bridge: Integrating WorkBuddy Models into DSH
The `dsh-workbuddy-bridge` plugin resolves the cumbersome configuration required for DeepSeek Harness to invoke local or specific service models, enabling seamless integration of WorkBuddy desktop app models and credits. It supports isolation between domestic and international versions, inference tier detection, credit display, and image input, and provides Web, Desktop, and TUI interfaces. Installation requires a logged-in WorkBuddy app and can be performed via the official source or source code. The plugin retains only an encrypted copy of credentials and is intended solely for personal learning and research.
Read MoreDSH: Continuing the Conversation — Syncing Sessions and Workspaces via GitHub Private Repositories
`dsh-chatsync` is an open-source plugin for DeepSeek Harness (DSH) designed to address issues with sessions, settings, and project files not being seamlessly carried over across multiple devices. The plugin stores runtime state in a GitHub private repository to enable cross-device synchronization. Installation requires an appropriate DSH version, Git, and the GitHub CLI. After logging in with the GitHub CLI, users connect the repository in DSH settings and select what content to synchronize. The first synchronization should be performed on the original computer. After that, a new computer can retrieve the data by connecting to the same repository. For daily use, it is recommended to synchronize before finishing work. The plugin supports both manual synchronization and automatic synchronization every 5 minutes. Note that a new computer must configure the API key separately. Dedicated credential files will not be uploaded, and it is recommended to keep local backups of important data.
Read More