DSH Plugin Gating Hub: Framework Upgrade Security and Plugin Version Gating
DSH Plugin Gating Hub is an upgrade-safety and plugin-management plugin designed for the DSH framework, intended to address version compatibility and supply chain risks. Its core features include one-click upgrades with automatic full-tree rollback, environment fingerprint detection to prevent accidental disabling, strict version gating via a semver engine, and a multi-source plugin marketplace (including private sources). This plugin ensures rapid recovery after failed upgrades, reducing administrators' trial-and-error costs; it must be installed and enabled manually.
Read Moredsh-auto-memory: A Claude Code–Style Automatic Memory Plugin for DeepSeek Harness
`dsh-auto-memory` is a plugin that provides persistent memory capabilities for DeepSeek Harness (DSH). It uses pure file-based storage to support typed memories, scope isolation, and data security mechanisms, addressing the issue of agents resetting memory in new sessions. The plugin automatically injects memory indexes and features atomic writes, truncation prevention, and zero external dependencies, making it suitable for developers who need to maintain long-term project context and coherence.
Read MoreDeepSeek Harness Plugin: Codex-Style Reasoning Level Slider
The official reasoning level options in DSH Harness (DSH) are fixed tiers; this plugin, `dsh-codex-effort-slider`, replaces them with a Codex-style continuous slider for more intuitive visual feedback. It is maintained by Microqian2th and licensed under the MIT license. Core features include support for Max/High/Low/Off, a blue-purple gradient, stardust effects at higher levels, light and dark theme compatibility, no runtime dependencies, and non-invasive behavior. The plugin supports only the DSH Desktop profile and requires a complete software restart after installation to take effect. It locates the DOM through official semantic attributes; if the official structure changes, it may stop working, but it includes a fail-open mechanism.
Read MoreScientific Figure Library: A Local-First Knowledge Base for Scientific Figures
Scientific Figure Library (SFL) is a local-first MCP server and application for scientific figures, designed to address the challenges of style reuse and code versioning in agent environments. This plugin allows users to store figures and code as immutable Releases on local disk and reuse them across multiple hosts, such as Pi, DeepSeek Harness, and Claude. SFL provides publishing, retrieval, previewing, and portable management features, supporting the local maintenance of a unified and traceable figure template library. It requires Node.js 22+.
Read MoreDeepSeek Harness Plugin: 550C Boot Animation
`dsh-550c-boot` is a playful reskin plugin for DeepSeek Harness, intended to recreate the boot process of 550C devices. By injecting the first frame and playing a full-screen animation, it overrides the native loading screen and unifies the desktop button color scheme into four phosphorescent styles. The plugin offers two modes—simple (about 4 seconds) and full (about 16 seconds)—and supports skipping at any time. After installation, restart the client. Version requirement: DSH `>=0.2.0-rc.1`.
Read Moredsh-wechat-plugin: Controlling DeepSeek Harness from WeChat
dsh-wechat-plugin is a plugin that enables bidirectional communication via WeChat between users and a DeepSeek Harness (DSH) agent running on a computer. It supports multimedia transfers, including text, files, and voice messages, as well as multi-session management, allowing users to control computer tasks through commands on their mobile device, confirm permissions, or receive push notifications. The plugin requires no public IP address, is open-sourced under the MIT License, and is primarily designed for developers who need to remotely control desktop agents. After installation, users can connect by scanning a QR code and use their phone as a terminal to efficiently manage DSH workflows.
Read Moredeepseek-harness-jev: Native DSH Plugin for Jev/Luna Decisions Integration
`deepseek-harness-jev` is a native DSH plugin that uses the Jev or Luna Decisions model to provide external assistance for DeepSeek Harness. It does not modify the host source code and integrates through the Cordis API, offering 12 features, including skill/file ranking, task supervision, instruction routing, and log cleanup. All features are disabled by default and must be manually enabled in the configuration page. The plugin is in an early stage, has limited compatibility, and its API judgments carry no correctness guarantees; when installing, ensure DSH version compatibility and manually enable it.
Read MoreStudyMate: A DSH Learning Assistant for Mathematics and Computer Science Subjects
StudyMate is an AI learning assistant for mathematics and computer science subjects, centered on the core philosophy of “learn with doing.” Maintained by Miaotofu01, it offers rich features, including course roadmaps, hands-on courseware exercises, real-time formula rendering, cross-subject memory, on-the-fly question generation and assessment, and customizable explanation styles. The plugin supports DSH learning mode and OpenAI ecosystem integration, helping users explore directions, overcome difficult concepts, and perform self-assessment through a systematic process. The StudyMate engine has been migrated to TypeScript, no longer depends on Python, and stores data locally by default.
Read MoreCC Safety Net: Pre-Execution Guard Plugin for AI Coding Agents
CC Safety Net is a plugin specifically designed for DeepSeek Harness, aimed at preventing AI agents from executing destructive commands or accessing sensitive information. It can intercept dangerous Git and file system operations, and block reads of SSH keys and environment variables. The plugin offers preset policies such as Standard and Strict, supports policy adjustments via a GUI, custom rulebooks, and policy version management through Git. Installation requires Node.js 18+. Note that it is not a sandboxing mechanism.
Read Moreanswer-me-with-html: Let Agents Answer Complex Questions with a Single HTML Page
The workflow plugin developed by QingYunA reduces model token consumption to 1/8 through a “Markdown draft + CLI processing” pipeline, generating well-structured HTML pages. It supports single-page output for complex use cases such as architecture diagrams and comparison tables, and offers offline availability, Always-on mode, and automatic layout. It requires Node.js 20+, and once installed via the CLI, it can be used across multiple Agent environments, addressing the limitations of traditional text output or inefficient HTML generation.
Read Moredsh-our-free-model: Free Model Plugin for DeepSeek Harness
The DeepSeek Harness plugin `dsh-our-free-model` is designed to address API key management challenges, offering a free solution for using frontier models—such as DeepSeek V4.1 Flash and Kimi K3—without registration or the need to enter an API key. This open-source plugin features over 14,000 stars, supports zero-dependency installation, and includes thirteen free access channels. Its core features include intelligent model selection, nine-page frosted-glass interface management, local gateway forwarding, and LAN access. Please note considerations for privacy and security: reasoning intensity control affects the output budget, and API keys are stored locally only.
Read MoreDeepSeek Harness Voice Control Plugin: Voice Input and Speech Output
The `dsh-voice-control` plugin leverages the Web Speech API to add voice interaction capabilities to the DeepSeek Harness Web interface. It supports voice input via the microphone icon or the `Ctrl+M` shortcut (with auto-send or draft insertion) and automatically reads aloud the assistant’s responses (up to 1,200 characters). The plugin includes a right-click menu for adjusting speech rate, selecting a voice, and previewing playback, and it runs entirely locally. Note that it only supports Chrome/Edge browsers, and speech recognition requires network access.
Read Moredsh-model-switch: Toggle Model and Thinking Strength with a Single Key Press
The DSH global plugin dsh-model-switch allows users to map any physical key (such as F13–F24 or a Bluetooth numpad) to a specific model and thinking intensity combination, addressing the inconvenience of frequent switching. It supports visual entry, real-time key monitoring, synchronization with the official model list, and includes built-in diagnostic tools. Once installed, users can trigger switching directly via peripheral keys without any pop-up interruptions. Currently, data is stored only in memory, with persistence planned for the future.
Read Moredsh-token-usage: Track Historical Token Usage of Local DSH Sessions
`dsh-token-usage` is a local plugin for DeepSeek Harness (DSH) designed to address the difficulty of tracking global Token usage. It features cross-session aggregation, global deduplication by ID, multi-dimensional grouping, and custom date range filtering. Data can be retrieved via the settings page, a local HTTP API, or command-line reports. It only counts assistant messages, has no extra dependencies, and helps developers accurately track costs.
Read Moregalvanize-dsh: Enable DSH Agent to Proactively Wake Up Conversations via Events
`galvanize-dsh` is a native plugin for DeepSeek Harness designed to replace traditional polling-based patterns with an event-driven architecture. It listens for events—such as files, emails, and webhooks—via the Galvanize core and automatically wakes headless DSH sessions to process tasks. The plugin provides tools such as `trigger_add`, `trigger_list`, `trigger_test`, and `trigger_remove`, and supports configuration through CLI commands and natural language instructions. Environmental requirements must be met before installation; after installation, the plugin automatically verifies connectivity and interacts with the core API to enable automated processing.
Read MoreDsh Image Annotate
`dsh-image-annotate` is a DSH plugin that allows users to annotate images directly within the editor. After installation, clicking an image opens the annotation dialog, which supports tools such as a brush, boxes, arrows, and editable text boxes. Text boxes can be moved, resized, edited, and deleted, with undo support included. Once annotation is complete, the generated PNG is inserted as a new attachment, leaving the original image unchanged. The plugin has no dependencies and a small footprint, requires restarting `dsh web` to take effect, and stores its configuration locally.
Read MoreDSH SuperTerminal: Turn the Side Terminal into a Workbench Beside the Conversation
DSH SuperTerminal is a plugin that extends Side Terminal into a fully featured local workbench, supporting multi-pane layouts, local agent management, and remote terminal connections. Users can perform basic tasks using natural language, form terminal discussion groups of up to six members to reach conclusions, or connect to remote servers via SSH. The plugin provides multiple theme colors, but it is currently in alpha and is subject to system restrictions and a cap on concurrent tasks.
Read Moredsh-session-folders: Add Folder Hierarchy to DeepSeek Harness Sessions
The `dsh-session-folders` plugin introduces standalone folder hierarchy management to DeepSeek Harness (DSH). It maintains a mapping between sessions and folders through a persistent registry, supporting session creation, renaming, archiving, and non-destructive removal. The plugin leverages `assignment[sessionId]` to achieve directory-based memory isolation, making it suitable for archiving sessions by project. Note that the current version does not support a tree-based grouping view in the sidebar.
Read Moredsh-restart-autoresume: Controlled Restart and Persistent Recovery
The Harzva dsh-restart-autoresume plugin for DeepSeek Harness (DSH) adds a controlled restart-and-resume task workflow to DSH. Relying on the existing `dsh-restart` service, it enhances robustness by managing a persistent ledger, refreshing Session state, verifying evidence, and automatically recovering crashed top-level Sessions. It supports selective resume, bounded concurrency, and configurable parameters. Installation requires meeting specific environment requirements and enabling `dsh-restart-autoresume` in the configuration file.
Read Moredsh-yzj: Yunzhijia x DeepSeek Harness Plugin
`guoxinshan/dsh-yzj` is the Yunzhijia integration plugin for DeepSeek Harness (DSH). It leverages the `yzj-cli` bridging system to provide 41 model-facing tools and a floating workspace panel. Core features include IM shell interactions compliant with Product Law v3.0 (with conversation-state and message-state switching), anchored @assistant replies in group chats, and support for sending to the assistant and replying in local threads. Installation requires running the relevant commands and restarting the GUI. Note that UI changes require rebuilding the bundle, and the Todo and AI Advancement features have been withdrawn.
Read Moredsh-plugin-guard: DeepSeek Harness Plugin Security Checkup
`dsh-plugin-guard` is a plugin security health-check tool for DeepSeek Harness, designed to identify risks in third-party plugins. It performs tiered plugin assessments through static code scanning, dependency review, AI-powered online auditing, and SafeSkill platform scanning. The tool can detect dangerous APIs, credential leaks, permission declaration mismatches, and supply chain risks, and presents results via a green/yellow/red dashboard. Plugins support version monitoring and alerts, helping developers identify potential security threats before introducing a plugin or during regular security inspections.
Read Moredsh-shipcheck: An Evidence-Driven Frontend Release Checker
`dsh-shipcheck` is a frontend release-checking plugin for DeepSeek Harness (DSH), designed to replace subjective judgment with evidence-first deterministic checks. The plugin uses a real browser to generate screenshots, named baselines, and JSON reports, covering accessibility, structure, content quality, and runtime error checks. It provides tools such as `shipcheck_run` and `shipcheck_matrix` for URL checks, structural regression comparisons, and user journey simulation. Acting as a release gate, it provides reproducible failure evidence and supports quality control before manual review, though it should be noted that it is structure-based rather than pixel-level.
Read Moredsh-where-am-i: A Neofetch Plugin for DeepSeek Harness
The `dsh-where-am-i` plugin addresses the lack of runtime environment awareness in models by injecting `neofetch`-style device information (such as OS, CPU, and memory) into session history in DeepSeek Harness. It uses native Node.js APIs, injects only on the first run, does not affect existing sessions, and skips subagents by default to avoid duplication. A service restart is required after installation. Its flexible configuration helps the model better understand the runtime environment.
Read MoreDeepSeek Harness: Minimal Voice Input Demo
The DeepSeek Harness Web UI Chinese Voice Input Plugin is designed to improve Chinese input efficiency in the Web interface. It relies on the browser's native speech recognition API. By clicking the microphone button on the right side of the input field, it converts speech into text in real time and appends it to the input field. After recognition, it does not automatically send the message, allowing users to edit it manually. Before use, the plugin must be installed via the terminal and the service must be restarted. Use cases include quick Chinese input and assisted note-taking. Note that it only supports the Web platform and the Chrome browser, and audio processing relies on browser services, so offline availability is not guaranteed.
Read MoreDeepSeek Harness Yoimiya Theme (dsh-yoimiya-theme)
`dsh-yoimiya-theme` is a fan-made theme for DeepSeek Harness (DSH), designed to provide an immersive development experience. Its core features include the “Summer Fest Night” and “Washi Day” palettes that switch automatically with the system, programmatic fireworks effects, smart wallpapers, warm-toned syntax highlighting, and a fully offline local music player. Installation requires DSH version 0.8.1+ and is supported on Web, command line, and Desktop; uninstalling will not delete music files.
Read More