dsh-synomega: Local Organic Reaction Prediction and Synthesis Planning

`dsh-synomega` is a local chemistry plugin developed specifically for DeepSeek Harness, addressing the cumbersome and error-prone issues associated with manually entering SMILES strings in traditional workflows. Built on SynOmega, it supports retrosynthetic analysis, forward synthesis prediction, route planning, feasibility scoring, and multi-component reaction evolution. All computations are performed locally and require no API key. Users can draw molecules with the Ketcher molecular editor, or in Web-enabled modes directly invoke tools through conversation to obtain reaction schemes and synthetic routes. Installation requires Node.js 22.19+ and Python 3.10+, and uses approximately 450MB of disk space.

Read More
Sakura-Afternoon Skin Plugin

Sakura-Afternoon is a visual novel-style theme plugin designed for DeepSeek Harness Web UI. By overriding default styles, it provides light and dark sakura palettes to create a subtle and elegant atmosphere. Core features include falling translucent petals, soft glow halos along the screen edges, header petal control buttons, and a master settings toggle. The plugin supports state persistence and automatically cleans up the interface upon uninstallation. Users must install it via the command line and enable it after restarting, through either the interface button or the settings menu.

Read More
DeepSeek Harness Plugin: request-frame-preview (Preview and Download Model Request Frames)

`request-frame-preview` is a DeepSeek Harness plugin designed to help developers inspect and debug the request parameters sent to a model. The plugin provides a “Request Preview” button in the session header, allowing users to view system prompts, tool definitions, and message flow details, and supports downloading the full request frame in JSON format. Data sources include the current running session or historical logs, and it strictly follows the DeepSeek adapter format. As a debugging tool, it exposes read-only HTTP endpoints, but note that its security is intended only for local debugging and should not be relied upon as a security boundary.

Read More
DeepSeek Harness Plugin Marketplace: Independent Team Directory and Brand Marketplace CLI

`dsh-plugin-hub` is a plugin marketplace designed specifically for DeepSeek Harness, intended to address pain points such as independent team catalog management, version publishing, and branded build workflows. Built on `pnpm`, it provides a release age policy and supports independent maintenance as well as the creation of branded marketplace CLIs. After verifying permissions and source code licensing, developers can visit its GitHub repository for installation and usage instructions.

Read More
z-asset/za_report: An Academic Writing Workflow Plugin for Asset Pricing and Machine Learning

z-asset/za_report is an academic writing plugin for DeepSeek Harness (DSH), designed for researchers in asset pricing and machine learning. It integrates literature review, data processing, model training, and report/PPT generation, automating research workflows through natural language commands. Licensed under the MIT License, it aims to improve researchers' efficiency by standardizing workflows.

Read More
z-asset/za_ppt: Generate Beamer/Quarto Presentation Slides from Academic Papers

z-asset/za_ppt is a plugin in the DeepSeek Harness (DSH) ecosystem, designed to reduce the time spent on report writing and slide preparation in asset pricing and machine learning research. It focuses on converting research papers into structured Beamer or Quarto slide decks. The plugin covers five core capabilities: literature review, data processing, model training, report writing, and slide preparation, and supports installation via dsh plugin or npm. Its domain calibration is centered on empirical asset pricing and machine learning. It depends on @deepseek-ai/cordis and dsh-skill and is released under the MIT License.

Read More
Z-Asset/ZA_literature: A Literature Review Tool for Asset Pricing and Machine Learning

DeepSeek Harness (DSH) has released the Z-Asset/ZA_literature plugin, designed to integrate key stages in the research workflow for empirical asset pricing and machine learning, including literature search, data processing, model training, report writing, and presentation creation. The plugin provides core capabilities such as literature review, data evaluation, model training, and document generation, and supports command-based installation with on-demand triggering within DSH sessions. It is purpose-built for research scenarios involving asset pricing, cross-sectional return forecasting, and deep learning, relies on a specific skill structure, runs through Cordis, and is open-sourced under the MIT License.

Read More
Z Research · Data Stage: A Data Discovery and Quality Assessment Plugin for Asset Pricing and Machine Learning

Z Research · Data Stage is a dedicated plugin within the DeepSeek Harness ecosystem for empirical asset pricing and machine learning research. It addresses the time-consuming challenges of data discovery and quality assessment through standardized workflows, supporting tasks such as data processing, model training, and report generation. Users can install it via CLI or npm and directly trigger its features using natural language instructions in a DSH session. The plugin relies on specific peer dependencies and is well suited to researchers who need to process financial and AI data efficiently, significantly reducing manual effort and operational costs.

Read More
@zasset/za-analysis: An End-to-End Analytics Plugin for Asset Pricing and Machine Learning Research

DeepSeek Harness (DSH) offers an end-to-end plugin named ZA Research Analysis, specifically designed for empirical asset pricing and machine learning research. The plugin integrates full-workflow automation for literature reviews, data processing, model training, and report/PPT generation. Users can install it via the command line and trigger relevant skills through natural-language instructions within DSH sessions, making it suitable for research scenarios such as factor models and cross-sectional return forecasting.

Read More
dsh-plugin-tool-repository: DSH Skills and Plugins Repository Browser

The plugin for DeepSeek Harness (DSH) introduces visual management of skills and plugins to the settings interface. Its core features include displaying skill lists by reading directory structure, supporting custom metadata and AI-generated bilingual descriptions; supporting enabling, disabling, and uninstalling plugins, with built-in plugins protected; and providing new-installation reminders, date maintenance, and alias input. Users can install the plugin via a command and, in “Settings → Repository,” browse details, generate descriptions, edit information, and manage plugin status.

Read More
dsh-wide-conversation: Customize the Conversation Column Width in the DSH Client

dsh-wide-conversation is a DSH plugin designed to address the restricted reading width of the conversation column on wide screens. It allows users to set the conversation width as a percentage of the available width (default: 100%). By overriding the built-in 680–920px constraint via CSS, it takes effect in the Web client without a build step and supports adjusting the `widthPercent` parameter through the configuration card.

Read More
dsh-plugin-attention-chime

The DeepSeek Harness (DSH) plugin `dsh-plugin-attention-chime` addresses the issue of missed requests during agent development caused by switching browser tabs. It leverages macOS notifications and sound cues to transform the scenarios of “waiting for a reply,” “asking a question,” and “requiring approval” into audible and visual signals. The plugin includes independent toggles and volume controls, a background task guard to prevent false positives, an escalation reminder mechanism with increasing volume, and the ability to click a notification to jump to the corresponding browser tab. Installation requires running specific commands and restarting the service, and it relies on `terminal-notifier` for jump-to-tab functionality. Configuration is flexible, supporting either setting pages or profile patches, and it runs locally with no external dependencies.

Read More
dsh-local-browser-search: A Local Browser-Driven Web Search and Scraping Plugin

The DeepSeek Harness plugin `dsh-local-browser-search` enables agents to directly control a local browser through Chrome DevTools Protocol, allowing search and web scraping without requiring an API key. Built on the Chromium engine, the plugin supports searches via Bing, Google, and others, and provides headless mode and timeout configuration. After installation, it registers the `browser_search` and `browser_fetch` tools in the agent. Each run uses a temporary profile to help ensure data privacy and security.

Read More
dsh-quick-window-launcher: Turning DeepSeek Harness Into a Standalone Application Window

This plugin opens DeepSeek Harness (DSH) in a standalone Chromium app window via a desktop shortcut, delivering a native app-like experience. It includes sidebar power controls, the `open_web` tool, and automatic version checking. Core features include desktop shortcut generation (using `--app` mode by default), watchdog-based process heartbeat self-healing and graceful restart, multi-signal liveness detection to avoid conflicts, and a pure Node.js helper script to ensure antivirus compatibility. After installation, Harness must be restarted, and options such as shortcut generation can be overridden through configuration.

Read More
dsh-opencode-go-sub: Injecting OpenCode Go Subscription and Zen Model Catalog into DeepSeek Harness

This plugin addresses the cumbersome process of managing OpenCode models in DeepSeek Harness (DSH). It registers the `opencode-go` (subscription) and `opencode-zen` (free) routes and, by dynamically fetching model lists, automatically discovering keys, and adapting to the new API, makes the DSH model selector resemble the official client. After installation, DSH supports zero-configuration key management and API adaptation. However, avoid conflicts with the built-in pi-ai provider and note the known limitations: image attachments are currently unsupported, and some models are subject to regional restrictions.

Read More
dsh-tool-adapt: Unofficial Model Compatibility and Security Adaptation Layer for DeepSeek Harness Web

`dsh-tool-adapt` is a compatibility and security adaptation plugin for the unofficial model families of DeepSeek Harness Web. Through core features such as dead-session protection, model-family convention injection, circuit breaking, and gateway compatibility diagnosis and repair, it addresses pain points including gateway compatibility errors, session deadlocks, and disordered permission management. After installation, a process restart is required. It supports hot configuration management via the settings panel and allows displaying status capsules to monitor operational status.

Read More
dsh-theme-cutout-clash: An Optional Neo-Brutalism Theme for DeepSeek Harness Web

The DeepSeek Harness theme plugin `dsh-theme-cutout-clash`, developed by YrracOwl, features a neo-brutalism and pop art style. It follows a "loud edges, quiet center" approach, using geometric lines and borders to decorate the interface while keeping the reading experience of conversational text smooth. It supports multiple variants, including Default, Cutout Clash, Night Cutout, and Pop Signal, and offers three motion preferences: full transitions, system-following, and transitions disabled. The plugin has no external resource dependencies, requires installation via the command line and a process restart, and should be used with preference coordination in mind when paired with the Acid Noir theme.

Read More
dsh-theme-acid-noir: DeepSeek Harness Web Dark Productivity Theme

dsh-theme-acid-noir is a non-invasive dark theme plugin for DeepSeek Harness Web. It provides two productivity themes, Acid Noir and Acid Noir Signal, supports integration with official settings, and does not modify the core source code. Users can install it with a command, switch themes after a restart, and must select a default appearance in settings before uninstalling. The plugin is compatible with DSH 0.1.2 and is suitable for developers seeking a cyber-style editing experience.

Read More
DeepSeek Harness Subagent Routing and Setup Plugin: dsh-subagent-conductor

`dsh-subagent-conductor` is an independent configuration and routing layer plugin that runs before the official subagent tool in DeepSeek Harness Web. It provides hierarchical routing policies at the session, role, and global levels, allowing users to precisely control the model provider, name, and inference mode for different root sessions in the input box, and supports bulk management of role templates through a visual interface or JSON. The plugin does not modify the DSH core, is compatible with versions 0.1.5 and 0.2, and depends on `@deepseek-ai/schemastery`.

Read More
dsh-mcp-pill: Lifecycle-Safe MCP State Capsule for DSH Web

dsh-mcp-pill is a client plugin for DeepSeek Harness (DSH) Web, designed to provide native visual status management for MCP connections. It offers input-box status pills, official-style settings cards, and a local RPC interface, supporting real-time monitoring of connection status, connection switching, and customizable pill placement. It is hidden by default and must be enabled in Settings. Installation requires DSH version 0.1.2 or higher, and the process must be restarted after installation.

Read More
dsh-better-input: DSH Web Input Optimization Plugin

dsh-better-input is a DSH plugin that adds an "AI Optimize Input" button to the left of the model selector in the web interface. The plugin supports custom prompt optimization, streaming backfill, undo operations, and hierarchical management of system prompts and appended prompts. Users can flexibly configure model parameters and prompts through a configuration file. Installation requires dsh 0.2.0-rc.2+ and the `web` profile as prerequisites, and runs `npm run link-deps` to resolve dependencies.

Read More
dsh-plugin-youcom: You.com Search and Scraping Plugin for DeepSeek Harness

The You.com search and fetch plugin for DeepSeek Harness extends an Agent's web access capabilities through the Cordis mechanism. It provides two providers, WebSearchProvider and WebFetchProvider, supporting search and web content extraction. Installation requires specifying a profile and configuring an API key, and users must explicitly specify the provider in the host configuration. The plugin has limitations, such as the fetch status code always being 200 and search results not including generated answers. It is suitable for agent development that integrates You.com as a search source in a DSH environment.

Read More
dsh-turnsnap: Zero-Config Per-Turn Git Checkpoints for DeepSeek Harness

The `dsh-turnsnap` plugin for DeepSeek Harness (DSH) automatically creates rollback-capable checkpoints for a Git working directory by monitoring agent turn events, addressing the problem of uncontrollable state in long sessions. It is ready to use after installation and requires no configuration. At the end of each turn, it automatically runs `git add -A` and commits, using a serialized queue to prevent concurrent conflicts. The plugin provides tools for pausing, resuming, and checking status, and can be controlled through natural language or scripts. Note that staged changes are merged, and checkpoint commits skip Pre-commit Hooks.

Read More
deepseek-harness-design-md-themes: 74 Native, Non-Intrusive Design MD Themes

DeepSeek Harness Design MD Themes is a non-intrusive official-style plugin designed to enable deep UI customization without modifying Harness core code. It includes 74 deterministic themes and provides rich, stable visual styles through native integration and persistent storage. Developers can install it via the DSH plugin manager, then select and apply a theme in Settings. It is compatible with the DeepSeek Harness 0.1.1-rc.2 environment.

Read More
dsh-nushell-only: Nushell-Exclusive Executor for DeepSeek Harness

The DeepSeek Harness plugin `dsh-nushell-only` takes over the shell execution layer to enforce that the model uses only Nushell. It introduces modifications across four levels: disabling the built-in sandbox and replacing it with a custom executor, preventing the model from switching to other shells, providing prompt-based instruction in the Nushell dialect, and intercepting and correcting common Bash/PowerShell syntax errors. It is compatible with DSH 0.2.0-rc.2 and is suitable for developers who want a unified shell ecosystem.

Read More