dsh-doc-index: Local Knowledge Base Index

`dsh-doc-index` is the local document index bundle for DeepSeek Harness, designed to enable semantic search over local documents. It supports Markdown, PDF, Office, and other formats, as well as Chinese word segmentation, through hybrid retrieval (BM25 + embedding vectors). The plugin supports incremental updates and precise line-number citations, and provides a command-line tool (such as `docindex query`) and a DSH service interface for model invocation. Installation requires Node.js 22.5+. By default, it uses a zero-dependency n-gram algorithm and can optionally use a Transformers model, with capacity limits and exclusion rules.

Read More
docgen: Documentation Workshop Skill Pack

DocGen is a documentation skills plugin for DeepSeek Harness, maintained by JohnXu22786. By using pure prompt-based Agent Skills, the plugin addresses the repetitive and time-consuming tasks of documentation writing, changelog organization, and code review. Its core capabilities include README generation, PR description generation, changelog generation, and code review. Users can invoke it through natural language commands, with support for Markdown output and style parameter adjustment. The plugin is dependency-free, open source, and secure, and is released under the MIT License.

Read More
dsh-db-connector: Secure Database Connector for DeepSeek Harness

`db-connector` is a standalone database connection plugin for DeepSeek Harness (DSH), supporting SQLite, PostgreSQL, and MySQL. It ensures data security through named connections, lazy opening, and credential reference management. Core features include schema introspection, read-only protection (blocking all statements other than SELECT), and a write approval mechanism. All operations generate JSONL audit logs. Installation requires Node.js >= 22.16, configuration supports YAML, and it registers three tools—`db_connect`, `db_schema`, and `db_query`—to manage connections, schemas, and queries.

Read More
DeepSeek Harness Plugin Introduction: context-pruner

The `context-pruner` plugin for DeepSeek Harness (DSH) is designed to address context bloat and token waste in long conversations. It uses heuristic rules to achieve deterministic processing with no model dependency, incorporating five independent context-screening mechanisms: handling outdated outputs, duplicate calls, failed calls, oversized blocks, and stale reasoning. The plugin supports configuring reserved regions and an exemption list, and it integrates seamlessly with DSH’s compression mechanism through structured audit reports. Developers can install the plugin from GitHub, via local mounting, or from a Git source, making it suitable for long-session scenarios that require token savings, such as logging and batch processing tasks.

Read More
Web Bridge — DSH Plugin: Real Browser Automation

Web Bridge is an MCP server plugin built on real browsers (Chromium/Firefox/WebKit) that provides agents with precise page manipulation capabilities through an accessibility tree snapshot mechanism. It supports navigation, clicking, form filling, screenshots, and script execution, and features tab management, multiple wait types, and session persistence. This tool is suitable for scenarios requiring high-precision interaction and dynamic web page processing, and can be deployed flexibly via configuration files or environment variables.

Read More
DeepSeek Harness Conversational Bookkeeping Plugin: Bookkeeping

The bookkeeping plugin for DeepSeek Harness (DSH) is designed to enable conversational bookkeeping through natural language. It accepts amounts and dates in natural language, including relative time expressions, automatically parses them, and stores the records in a local SQLite database. It supports multiple currency formats and custom category rules. Its features include income and expense queries, monthly budget management, report export in CSV or HTML format, and data persistence. The plugin offers both a conversational mode and a CLI mode, and can be integrated into a DSH process or run standalone, making it suitable for developers who need an agent to manage personal financial records.

Read More
DeepSeek Harness Plugin: `read_document` Tool

The `@jiaoqsh/dsh-document` plugin for DeepSeek Harness (DSH) enables models to directly read documents in multiple formats locally. It supports formats such as Word, PowerPoint, Excel, PDF, and RTF, converting files into Markdown for model parsing without the need for external APIs or network access. It supports pagination, metadata extraction, and file system integration. Users must install the plugin by specifying a profile and configure parameters (such as file size limits and the number of PDF pages). The plugin runs locally but does not support OCR and has specific limitations in handling PDF layouts.

Read More
DeepSeek Harness Billing Plugin: dsh-token-billing

DeepSeek Harness provides a plugin called `dsh-token-billing`, designed to support cost monitoring during model inference. The plugin displays billing readouts directly below the Web UI input box, supports total cost aggregation by session and workspace, and includes peak and off-peak time-based pricing. Data is persistently stored in the `token_billing` domain, with detailed billing breakdowns available on hover. Installation is simple, requiring only a single command, and the plugin can be disabled through configuration. When using it, users should pay attention to model pricing rules and source-code review.

Read More
dsh-theme-blackhole: DeepSeek Harness Theme Plugin

`dsh-theme-blackhole` is a theme plugin designed for the DeepSeek Harness (DSH) Web UI, providing a visually striking sci-fi aesthetic. The theme uses WebGL to render a Schwarzschild black hole background through null-geodesic ray tracing, features deep-space glass panels, and incorporates amber accent colors. It supports first-class theme management, bidirectional synchronization with appearance settings, and persistent state, along with performance-friendly fallbacks such as a 30 FPS cap. Installation requires Web service permissions and is only supported in profiles that include a web server. It is open source under the MIT license.

Read More
dsh-skills-manager: Managing Skills in the DSH Settings Panel

dsh-skills-manager is a DSH client plugin that provides a visual settings panel for skill management. It supports searching, pagination, enabling/disabling, and deleting skills with a Recycle Bin feature. Built-in skills are read-only protected, and all operations are restricted by security boundaries. With this plugin, users can manage skills flexibly without uninstalling, effectively reducing the risk of accidental misuse.

Read More
dsh-auth-gate: An Authentication Gate Plugin for DeepSeek Harness

`dsh-auth-gate` is a security plugin for DeepSeek Harness (DSH), designed to provide login authentication and protection for the Web UI. It intercepts unauthenticated requests upfront through the Nginx `auth_request` interface and offers core features such as SVG image CAPTCHA, brute-force protection (with IP and account rate limiting), single sign-on, and encrypted password storage. The plugin listens on localhost by default and must be configured through an Nginx reverse proxy. After the first login, changing the default credentials is mandatory. To ensure security, public-facing deployments should use HTTPS.

Read More
dsh-dejavue: DeepSeek Harness Plugin Duplicate Checker

`dsh-dejaview` is a plugin for DeepSeek Harness, designed to address the problem of developers reinventing the wheel when building new plugins. The plugin registers the `check_plugin_novelty` tool, which retrieves data from the awesome-dsh-plugin registry, GitHub `dsh-plugin` topics, and npm packages, and uses IDF-weighted semantic similarity to score and verify the novelty of plugin ideas. Before building a plugin, an Agent can invoke this tool to obtain candidate plugins, similarity scores, and decision guidance, helping the model make innovation judgments based on objective data and improve development efficiency.

Read More
DeepSeek Harness Built-in Debug Console Plugin

`dsh-sseye` is a debugging plugin for DeepSeek Harness that provides proxy-free observability into internal LLM calls. By intercepting the `llm/stream` layer, it captures full call-chain data, including system prompts, streaming responses, and tool calls, and presents it in a DevTools-style interface. The plugin supports real-time comparison, JSON export, and regex-based masking of sensitive information, while data is stored in memory only by default to ensure privacy and security. Once installed, you can view detailed token counts, latency, and reasoning processes in the interface, making it easier for developers to troubleshoot LLM invocation issues.

Read More
DeepSeek Harness Audio Notification Plugin: Full-Turn Response and Prompt Reminders

The DSH Sound Plugin for DeepSeek Harness (DSH) synthesizes audio cues using the Web Audio API to provide sound reminders when an agent finishes work or awaits input. Core features include: a “ta-da” sound when a full response ends, a “ding” sound while awaiting input for a question, and triggers only when background tasks truly complete to avoid false alerts. It supports flexible control over muting, volume, and ringing only when the tab is hidden via localStorage. With no external dependencies, no required external files, simple installation, enabled by default, and easy configuration, it effectively addresses the pain point of developers not being able to constantly monitor the interface while tasks run in the background.

Read More
jedzqer/dsh-retry-plugin: DSH Auto-Retry Plugin

The plugin-based ecosystem of DeepSeek Harness (DSH) aims to lower the barrier to development. To address task interruptions caused by failed AI API requests due to network issues or rate limits, the community has contributed the `jedzqer/dsh-retry-plugin`. When a request fails, this plugin automatically sends a “continue” message to retry, improving workflow stability. Public documentation currently does not provide official installation commands or detailed usage; see the GitHub repository for details. When installing, pay attention to plugin permissions and license information.

Read More
dsh-plugin-git-status: DeepSeek Harness Workspace Git Status Panel and Tools

dsh-plugin-git-status is a DeepSeek Harness (DSH) plugin designed to visualize Git status and integrate it into the editor. It provides status chips in the editor tool line, an expandable change list, and a `git_status` tool. The plugin operates in read-only mode, auto-refreshes every 30 seconds by default, supports configurable refresh intervals, and pauses refresh when the browser tab is hidden. After installation, the DSH process must be restarted. It is suitable for scenarios that require monitoring code changes or accessing Git information.

Read More
dsh-console: DeepSeek Harness Console Plugin

`dsh-console` is a plugin for DeepSeek Harness (DSH) that integrates Web service management, SSH tunnel operations, and console functionality into the command-line plane. Installation requires specifying `--profile web`. Core features include managing Web UI state, establishing SSH tunnels, launching the TUI interface, and sending a single query. Users must configure the SSH key path, local/remote ports, and TUI startup command in `cordis.patch.yml`. Note that stopping the Web service terminates the process, and ensure SSH keys are configured only for trusted sources.

Read More
dsh-desktop: A Plugin for Directly Launching the Native DSH Desktop Client

`dsh-desktop` is a DSH plugin designed to deliver a native desktop application experience using Electron. It uses IPC communication instead of HTTP services, enabling one-click window startup without manually opening a browser while reusing all of DSH's web capabilities. The plugin supports hiding the native title bar, customizing the UI style, and a two-way cleanup mechanism when the window is closed. To install it, download the Release package and link it to a DSH profile, then launch it with `dsh --profile dsh-desktop`; terminal behavior can be controlled via environment variables.

Read More
dsh-model-slider: Enhanced DeepSeek Harness Model Selector

The `dsh-model-slider` plugin optimizes the model-switching workflow in DSH Web UI through slider and panel integration. It provides low/medium/high quick-switch tiers, advanced selection, multimodal icons, and provider balance monitoring. It takes effect after installation, with its configuration file located at `~/.dsh/.model-slider.json`; it supports membership and prepaid usage queries and securely reads API keys.

Read More
DSH Lifeboat: A Process-Free Secure Boot, Fault Isolation, and Recovery Tool for DSH Configuration Files

DSH Lifeboat is a DSH plugin maintained by IoveCelestina, designed to address the diagnostic challenges caused by Harness Loader crashes or Profile configuration failures. It uses a standalone Rescue server to isolate and recover from faults. Its core features include dual-mode interaction via Web UI and CLI, intelligent diagnostics based on snapshot cloning and an "upper-bound-first removal" algorithm, and full fingerprint verification with an atomic recovery mechanism. The plugin remains accessible when the main process crashes, supports one-click rollback, and can generate JSON-formatted diagnostic reports. After installation, the diagnostic service can be enabled. It requires a Node.js environment and provides a reliable fallback for Harness Profile.

Read More
DeepSeek Harness (DSH) Plugin Sorter (`dsh-plugin-sorter`)

`dsh-plugin-sorter` is a plugin management tool for DeepSeek Harness (DSH), designed to simplify configuration management as the number of plugins grows. It provides a two-column drag-and-drop interface, supporting plugin enable/disable, load order adjustment, and viewing details and underlying Loader information. The tool relies on an ESM build and includes a draft-saving feature to prevent accidental changes. Users can install it via CLI, manage it from the settings page, and restart DSH after manually saving to apply the changes.

Read More
DeepSeek Harness Plugin: project-ai-docs

`project-ai-docs` is a DeepSeek Harness (DSH) plugin designed to address agent session context limits and cognitive fragmentation. It enables knowledge consolidation by maintaining a project-specific `docs/.ai/` knowledge base. The plugin uses map-first, layered budgets, and closed-loop write-back mechanisms to ensure structured documentation and incremental awareness. Users can install it through the official DSH channel or copy the skills directory to use it in other hosts, following an initialization, task execution, and write-back workflow to make project knowledge persistent and verifiable.

Read More
Maglev for DSH: An Introduction to the DeepSeek Harness Plugin

Maglev for DSH is a standalone plugin for DeepSeek Harness (dsh), designed to address dsh’s shortcomings in software development process semantics. With 29 built-in skills, Maglev GUI (Truth Cards and Crystallization Cards), and a spec completeness validation gate, the plugin adds process discipline and knowledge accumulation mechanisms to dsh, helping teams achieve long-term, controllable project iteration. After installation, no manual configuration is required: skills are automatically injected, and tools can be invoked directly in dsh to read the current state and crystallize conclusions.

Read More
dsh-provider-auto: DeepSeek Harness Inference Model Auto-Configuration Plugin

`dsh-provider-auto` is a DeepSeek Harness (DSH) plugin designed to address the issue where custom OpenAI-compatible gateways cannot display reasoning controls in the frontend. It identifies models by name (e.g., `gpt-5*`, `o1*`, etc.) and automatically injects the `reasoningEfforts` configuration for eligible models, supporting four levels: off, low, medium, and high. The plugin supports protocol listening and declaration retention, and default rules can be overridden via a configuration file. It requires a Node.js 22+ environment.

Read More
dsh-plugin-scout: DeepSeek Harness Ecosystem Reconnaissance and Assessment Tool

dsh-plugin-scout is a reconnaissance plugin for the DeepSeek Harness ecosystem, designed to scan core repositories and community repositories tagged with dsh-plugin to identify existing resources and assess their feasibility. It supports full-ecosystem scanning, list generation, relevance filtering, and code review, and can output TRY/WATCH/SKIP decisions. Developers can use it before developing a new Harness to quickly evaluate community solutions. Note that its responsibilities are limited to reconnaissance and evaluation; installations require checking the source code and license.

Read More