dsh-live-loop: A Frontend Runtime Verification Plugin for DeepSeek Harness
`dsh-live-loop` is a frontend runtime validation plugin for DeepSeek Harness (DSH), designed to ensure that web applications load correctly and behave as expected by performing detection, interaction, visual comparison, and evidence generation. It does not modify code and is responsible only for the validation process, supporting frameworks such as Vite, React, and Vue while providing isolated environments, bounded observation, and a strict four-state verdict mechanism. It requires DSH version 0.1.0-rc.7 and Node.js 22.19.0+, giving agents a closed-loop capability from detection and remediation to validation to ensure rigor in frontend automation workflows.
Read Moredsh-bull-bear: DeepSeek Harness Real-Time Market Pet
`dsh-bull-bear` is a market pet plugin for DeepSeek Harness that combines real-time A-share data with anthropomorphic animations to visually present portfolio sentiment. The pet runs or jumps based on the portfolio’s average price change (above 5% is bullish; below -5% is bearish), and its speed increases as returns rise. The plugin supports dragging, holdings management (up to 200 positions), and a market race view. Data is sourced from the Sina Finance API and is provided for learning and research purposes only; it does not constitute investment advice. The installation command is `dsh plugin --profile web add dsh-bull-bear`.
Read Moredsh-http-probe: HTTP Probe Tool for DeepSeek Harness
The DeepSeek Harness (DSH) plugin `dsh-http-probe` provides the Agent with lightweight HTTP probing capabilities. Using the `http_probe` tool, it allows the Agent to retrieve only the status code, latency, and response headers of an HTTP(S) endpoint without downloading the full response body, making it ideal for API health checks. Since the plugin is not published to npm, it must be installed directly from a GitHub URL. Once installed, the Agent can directly use this tool to probe the connectivity of any endpoint.
Read Moredsh-suggest-reply: Sidebar Candidate Reply Generation Tool
`dsh-suggest-reply` is a sidebar plugin based on DeepSeek Harness, designed to help users generate candidate lines for AI replies. Users define tasks and JSON format through a custom System Prompt; the plugin automatically reads the latest message and optional director requirements to generate a style list. The generation process is isolated from the main session and supports real-time streaming debugging, parameter configuration, and one-click insertion of results. The plugin depends on `dsh-better-sidebar`, providing a clean, standalone generation experience.
Read Moredsh-schedule-command: In-Session Scheduled Task Management Plugin
`dsh-schedule-command` is a DeepSeek Harness plugin that allows users to create and manage local scheduled tasks in conversations via the `/schedule` command. Maintained by user piaohua, the plugin supports defining one-time or recurring execution plans in natural language and automatically mounts its runtime upon installation. Users can view and delete tasks, but tasks only trigger while the session is active; after sleep recovery, only the latest missed run is replayed. In addition, the underlying v1 implementation does not support Cron, so complex calendar rules fall back to fixed intervals.
Read Moredsh-shop-assistant: DeepSeek Harness E-commerce Workflow Plugin
dsh-shop-assistant is a DeepSeek Harness workflow plugin designed specifically for e-commerce scenarios. It streamlines tedious tasks such as responding to negative reviews, generating competitor copy, and calculating profit. The plugin can read CSV review data and competitor page information, combine them with a store policy knowledge base to generate compliant responses and marketing copy, and perform reproducible profit margin calculations using fixed formulas. To use it, developers must first install DeepSeek Harness, install the plugin via the command line, and place relevant files (such as reviews.csv) into the workspace folder. The plugin is released under the MIT license and supports Chinese-language environments.
Read Moredsh-right-drawer: DeepSeek Harness Right Drawer Host
The DeepSeek Harness plugin `dsh-right-drawer` provides a host solution for the right-side panel on the Web. It manages content plugins through the `dockPanels` registry, supports full-height drawer display, and features automatic tab stacking, drag-to-adjust width (280–900px), and coordinated layout with the scheme panel. Adopting a plugin-based architecture, content plugins can be integrated simply by registration, making it suitable for developers who need to build complex sidebar layouts.
Read MoreIntroducing DSH Plugins: Agent Shell and the Shell Drawer
`dsh-agent-shell` is a client plugin for DeepSeek Harness that provides command execution and visual feedback. It registers the `bash`, `cmd`, and `shell_drawer` tools and uses a loopback RPC channel to push execution results in real time to the Shell drawer on the right, supporting timeouts, output truncation, and manual control. Installation requires a dependency on `dsh-right-drawer`. It enables system commands to be executed through conversation or tool instructions, making it suitable for automated testing and debugging scenarios.
Read MoreDeepSeek Harness Plugin dsh-aidd: An AI-Driven Drug Discovery Research Workbench
`dsh-aidd` is an AIDD research workbench plugin built on DeepSeek Harness, designed to address challenges of scientific rigor and data traceability through structured workflows. The core of the plugin includes a ReAct research-cycle engine, versioned artifact management with provenance information, and 17 research skills (including AIDD-specific skills such as virtual screening and molecular generation). Installation supports two modes: Profile Bundle and Agent Preset, providing end-to-end workflows from initialization and experiment execution to reproducible modeling. The project emphasizes data standards and reproducibility, with built-in data-discipline guardrails and automated testing to ensure stability.
Read Moredsh-worktree-studio: DeepSeek Harness Manual Git Worktree Task Board
`dsh-worktree-studio` is a manual Git worktree task board plugin for DeepSeek Harness. It maps Git branches, worktrees, and DSH sessions one-to-one, supporting the creation of isolated environments, native session integration, state validation, and merge previews. Designed to be non-invasive and without consuming model tokens, it aims to enable isolated verification and safe merging of code changes locally through Git worktrees and a state token mechanism.
Read Moredsh-trace-repeat: Task Trace Recording and Replay Plugin
The `dsh-trace-repeat` plugin provides precise execution snapshots for agent development in the DSH Web environment through atomic versioning. It sequentially records immutable versions of three types—reasoning, tool, and user—along with full context metadata. By leveraging a built-in Git repository, the plugin supports materializing historical versions as worktrees and resuming execution from the “Reasoning Completed” version using the `/trace resume` command. It requires no configuration, automatically records sessions, and helps developers trace and reproduce the execution state at any point in time.
Read Moredsh-env-switcher: A Plugin for Switching Between Windows and WSL2 DSH Instances
The `dsh-env-switcher` plugin is designed to address the cumbersome switching process when maintaining DeepSeek Harness (DSH) instances across both Windows and WSL2 environments. It allows users to switch directly via the "Runtime Environment" tab in the Web UI, eliminating the need to manually copy and paste URLs. The plugin supports coexistence mode (with default ports 3080 and 3081) and features such as sidebar-collapse awareness and independent session history. Users can install it via the official command or from source, and configure the port and path in `cordis.patch.yml`. After installation, restart the server to start using it.
Read Moredsh-plugin-mgr: DeepSeek Harness Plugin Manager
`dsh-plugin-mgr` is the plugin management plugin for DeepSeek Harness (DSH), designed to manage plugins through a web-based card interface. Its core features include displaying plugins in card format, supporting enable/disable, uninstall, one-click updates (npm registry only), and search filtering; it uses a Host/Client architecture for configuration read/write and UI rendering. The plugin includes security restrictions (managing host modules is prohibited), concurrency protection, and immediate error display, effectively addressing the complexity of dependency management and status checking.
Read Moredsh-model-health: DSH Model Availability and Latency Testing Plugin
The `dsh-model-health` plugin for DeepSeek Harness (DSH) is used to manage the availability and performance of multi-model configurations. It provides a Web UI panel to display model details and supports concurrent testing of all models (up to 6 models, with a 10-second timeout), with results stored persistently. Additionally, the plugin provides a `list_models` tool for viewing the model list during a conversation. Installation requires specifying dependency versions.
Read MoreDeepSeek Harness (DSH) Tavily Web Search Plugin
The DeepSeek Harness (DSH) plugin `dsh-tavily-search` decouples search from the LLM by connecting directly to the Tavily Search API and consuming only Tavily quota. Maintained by ouones, it supports installation and automatic mounting. Before use, install the DSH environment (Node.js 22.19+) and obtain an API key. It is recommended to store the key in the DSH credential service; default settings can be adjusted as needed in `cordis.patch.yml`. After restarting, verify the setup with `dsh --dump-config`. The plugin is implemented purely in JS and is licensed under the MIT license.
Read Moreoukeming64-tech/codex-skills: Evidence-First Agent Skills for Handoff Auditing and Documentation
`codex-skills` is a community-developed, instruction-based skill set for DeepSeek Harness, designed to address edge cases in which code, evidence, and documentation become inconsistent during agent handoffs. It includes two core skills: `handoff-auditor`, which audits stated claims against the actual state before a handoff; and `docs-sync-guardian`, which enforces synchronization between documentation and code changes. The plugin is lightweight, delivered solely via the file system, and depends on neither network access nor background services, making it suitable for validating others’ handoffs or preventing documentation drift.
Read Moreostar-dsh-left-sidebar: DSH Left Sidebar Workspace Manager
The DeepSeek Harness (DSH) plugin `ostar-dsh-left-sidebar` enhances management capabilities on top of the official left sidebar. It supports bulk deletion, thorough data cleanup, session migration and copying, grouped favorites, drag-and-drop reordering, and locating the current session, while faithfully replicating the official style. Installation requires meeting the Node.js and pnpm version requirements; install it via the command line and restart. The plugin has no build step, uninstalling it does not affect data, but the browser cache must be manually cleared to remove the grouping memory.
Read MoreDeepAtlas (dsh-deepatlas) — DeepSeek Harness Task-Aware Plugin Navigation
DeepAtlas is a task-aware plugin navigation and auditing tool for DeepSeek Harness (DSH), aimed at mitigating plugin risks arising from the large DSH ecosystem. It scans the GitHub ecosystem, recommends plugins based on task requirements, and audits them across multiple dimensions, including provenance, code, and dependencies, while supporting local storage and controlled installation. DeepAtlas is in Public Preview and requires DSH version 0.1.1-rc and Node.js. Users can use the `deepatlas` command suite to complete the full workflow from status checking, scanning, and retrieval to auditing and confirmed installation. The tool emphasizes permission security: index and audit data are stored locally, and installation is executed only after explicit user authorization, ensuring evidence-based decision-making for plugin usage.
Read Moredsh-mem0: Self-Hosted mem0 Memory Read/Write Plugin
`dsh-mem0` is a plugin for DeepSeek Harness (DSH) designed to manage an agent's independent long-term memory. It integrates with a self-hosted mem0 API service and provides tools such as `mem0_add`, `mem0_search`, and `mem0_update`, while supporting Dashboard authentication and a hot-swap mechanism. The plugin requires no modifications to the DSH source code, is configured via a browser-based configuration card, and securely manages memory data, making it suitable for scenarios that require persistent storage of agent context.
Read MoreDeepSeek Harness Projection Shard Warm-up Plugin
The DSH projection plugin is designed to address event loop blocking and timeout issues caused by cold folding of large sessions. Its core mechanisms include pre-warming chunks to yield the event loop, cache backfill from forked child sessions, and background line completion for disk-cold sessions. These features reduce cold folding time from minutes to milliseconds and provide statistics and configuration APIs. Users should be mindful of synchronous decoding limits, memory risks, and compatibility requirements when using the plugin.
Read Moredsh-fork-perf: DSH Conversation Fork Performance Optimization Plugin
The `dsh-fork-perf` plugin for DeepSeek Harness (DSH) is designed to address performance bottlenecks when forking large sessions, reducing the time taken by approximately 1/18. It reuses references to frozen JSON trees through a "zero-copy fork" and optimizes the initialization process of `PersistenceCoordinator.initFor`. The plugin provides an HTTP API for monitoring and configuration, along with robust fallback mechanisms and patch restoration capabilities. Note that it depends on DSH version 0.1.0-rc.6, and is limited by the fact that copies inside the `enqueue` closure cannot be eliminated.
Read Moredsh-plugin-anti-rot: DeepSeek Harness plugin for compressing tool error payloads
The DeepSeek Harness (DSH) plugin mitigates “context rot” in long sessions by compressing verbose tool error stack traces. In the `tools/post-execute` phase, it identifies failure signals (`isError` or a non-zero exit code), truncates error text to a specified number of lines and removes indented stack lines, while preserving a truncation marker. After installation, it must be enabled in the configuration file with a positive integer `maxLines` parameter. The tool effectively reduces noise and saves context budget, leaving successful calls unchanged.
Read MoreDeepSeek Harness Web Authentication Plugin dsh-auth
`dsh-auth` is an authentication gateway plugin specifically designed for the DeepSeek Harness (DSH) Web interface. It enhances system security by supporting password-based login, automatic logout after idle periods, session expiration, and single sign-on. Users can configure credentials and expiration times via the command line or the Web interface. The plugin also supports login history auditing and reverse proxy access. After installation, the default password must be changed immediately, and HTTPS is recommended for production environments. This plugin provides essential security hardening capabilities for the DSH Web interface.
Read Moredsh-kernel-mesh: DeepSeek Harness Multi-Kernel Model Routing Plugin
`dsh-kernel-mesh` is a kernel mesh plugin for DeepSeek Harness (DSH) designed to enable seamless switching across multi-model collaboration. It packages the Kimi, Grok, Codex, and MiniMax models as plugins, providing model routing, subagent recipes, and management tools that support checking status, dispatching tasks, and switching routing paths. Installation requires adding the plugin itself and each model kernel dependency separately, along with manually configuring agent presets. The plugin supports fallback routing and vendor-specific search tools, with each kernel having its own independent system prompt. With this plugin, users can uniformly manage multi-vendor models and subagents within DSH.
Read MoreIntroduction to the DeepSeek Harness Voice Input Plugin
`dsh-voice-input` is a local speech-to-text plugin for DeepSeek Harness. Built on FunAudioLLM's SenseVoice model, it integrates microphone support into the Web Profile to automatically transcribe recordings. It does not upload audio to the cloud, helping protect privacy. Installation requires Python 3.10+ and Node.js >= 22.19.0. The plugin supports one-click recording, automatic model download management, and background service startup. The first use requires waiting for the model download, and the project is released under the MIT open-source license.
Read More