dsh-suggest-prompt: DSH Web's next prompt suggestion plugin
dsh-suggest-prompt is an MIT-licensed plugin for the DeepSeek Harness (DSH) web interface, designed to address the pain point where users need to re-enter prompts after the end of an agent turn. After each agent turn completes, the plugin generates the next suggested prompt through a bounded auxiliary LLM call. The generation process inherits the main request routing by default, sends the content of the last round of conversation, and disables thinking mode to optimize performance. In the web interface, suggestions are displayed as light-colored ghost placeholder text within the input box and are only visible when the draft is empty. By default, users can accept suggestions into the draft using the Tab key, or configure shortcuts such as Alt+Slash or Ctrl+Enter. The plugin features strict security and quality filtering mechanisms, including transcription desensitization and output net...
Read Moredsh-crew: Execute changes in dsh using PM and Role Agents
dsh-crew is an MIT-licensed plugin for DeepSeek Harness (dsh) designed to transform dsh sessions into Product Managers (PMs) by orchestrating a set of role-playing agents (such as architects, engineers, QA, reviewers, etc.) to collaborate on agent workflows. Roles exchange information via disk files, and users only need to interact with the PM. The core functions are divided into two lanes: the "Ask lane" for questions and status queries, which does not generate code changes; and the "Team lane" for executing changes, following the "one change, one task" principle. The process includes a round of QA, various reviews, and a final commit. All decisions are recorded in repository documentation, while push and publish operations must be manually confirmed by the user each time. The plugin requires Node.js 18 or higher and supports via...
Read Moredsh-ui-concise: DeepSeek Harness WebUI's Concise Session Mode
dsh-ui-concise is a session display plugin for the DeepSeek Harness WebUI, maintained by stars2022 and released under the MIT License. This plugin aims to provide a concise session mode similar to the Codex/macOS style, organizing Turns, tool calls, thinking content, and final answers into a more compact view. Core features include: the settings page provides "Concise Mode" and "Hide Thinking Content" options; each Turn features an independent timer that counts down in real-time while active, automatically collapses upon completion, and remains expanded in case of an error; consecutive tool calls are merged into an activity bar, displaying the status while running, a summary upon completion, and supporting expansion to view original details; the final Assistant answer is retained outside the collapsible section to maintain normal Markdown formatting and operation styles; enabling "Hide Thinking Content".
Read Moredsh-hyperframes: Register HyperFrames by HeyGen's 20 video skills into DSH
`dsh-hyperframes` is a video creation skill plugin for DSH (DeepSeek Harness). It aims to integrate HyperFrames by HeyGen's official 20 HTML video generation and editing skills into agents through the "everything is a plugin" mechanism, avoiding redundant development of rendering workflows. The plugin syncs skills including core concepts, animation adaptation, audio response, embedded subtitles, command-line tools, and various scenario-based video production capabilities (such as product launches, music-to-video, Figma integration, etc.), based on HeyGen v0.8.20. Users can quickly install and restart the service via the `dsh plugin` command, and then directly describe requirements in the conversation (e.g., "convert a URL into a video") or execute `npx hyperfram...`
Read Moredsh-docker: DSH's Docker container management plugin
dsh-docker is a container management plugin for DSH, designed to enable agents to securely operate local Docker objects. The plugin invokes the Docker CLI via the official subprocess service, uses an argv array to pass parameters to mitigate shell injection risks, and retains the default approval mechanism in `docker_exec`. It provides six core tools: `docker_ps` (list containers), `docker_images` (list images), `docker_logs` (view logs), `docker_inspect` (view container details), `docker_exec` (execute commands inside the container), and `docker_manage` (lifecycle management, such as start, stop, restart, and delete). The plugin requires Node.js.
Read Moredsh-dingtalk: DeepSeek Harness DingTalk Group One-way Notification Plugin
dsh-dingtalk is a community plugin for DeepSeek Harness designed to simplify the process of sending notifications to DingTalk groups after agent tasks are completed. Built on pure Node.js with no runtime dependencies, it supports Windows, macOS, and Linux across all platforms. It encapsulates the DingTalk custom bot webhook, providing two tools: `dingtalk_notify` (for sending Markdown messages) and `dingtalk_text` (for sending plain text messages), and supports signature security mode. Users can install it via the `dsh plugin` command and configure `webhook`, `secret`, and `timeout` in `cordis.patch.yml`.
Read Moredsh-calendar: Connect CalDAV calendar tools to DSH
`dsh-calendar` is a calendar plugin for the DSH community that provides scheduling capabilities to agents via the CalDAV protocol. It encapsulates five model tools: list, create, update, delete, and search, supporting Google, iCloud, Nextcloud, and any custom CalDAV server. Licensed under the MIT license and requiring Node.js >= 22, the plugin uses Basic authentication and mandates the use of application-specific passwords. It does not support OAuth or a web settings page; configuration must be completed via the `cordis.patch.yml` file or environment variables. When used, users must configure corresponding parameters based on the service provider type, such as the `calendarId` for Google or the full URL for iCloud, allowing mainland China users to configure [it].
Read Moreoctie-dsh-plugin: State-oriented task graph kernel in the form of a DSH plugin
octie-dsh-plugin is a DeepSeek Harness (DSH) bundle plugin maintained by StarChen-Cycler, encapsulating the octie state-oriented task graph kernel as an installable component. Its core philosophy is "state derivation over manual setup"; it constructs task graphs through atomic tasks and dependencies, automatically unlocks downstream tasks via BFS upon completion, and supports resuming long sessions with immutable snapshots. The plugin offers a standalone CLI, Web UI, and DSH interface that share the same task graph storage, featuring 13 model tools, Cordis service, and event support. The CLI and Web UI support terminals, scripts, and visual DAG views, facilitating the planning, execution, and tracking of tasks within DSH sessions. Installation requires Node.js ≥ 20 and is recommended via npm or GitHub, but GitHub...
Read Moredsh-plugin-rollout-scout: Scout the current account's dialogue model in DSH
dsh-plugin-rollout-scout is a DSH plugin maintained by SpookySandwich designed to detect the current conversation model version assigned to the account. The tool initiates temporary probing sessions, reads reasoning-delta data during chain-of-thought streaming output, and utilizes a heuristic scoring mechanism based on paragraph opening features to distinguish between new and old models. If determined to be an old model, the plugin aborts the probing session as soon as possible to conserve resources; if determined to be a gray-scale or new model, it allows the session to complete and offers the option to retain it. The plugin supports configuring probing prompts, concurrency levels, scoring thresholds, and early stopping parameters such as TPS/TTFT, and features a sidebar console for managing session operations like pausing, resuming, forcing stop, and cleaning up. Users can hover to pin specific sessions and mark them for retention. The plugin utilizes ownership records to prevent the accidental deletion of non-probing sessions.
Read MoreMindspace DSH Session Memory Plugin: Adding Editable Multi-user Session Memory for DeepSeek Harness
`mindspace-dsh-session-memory` is a community plugin for DeepSeek Harness (DSH) designed to provide editable, session-isolated multi-user memory capabilities. To address the issue of context confusion in long sessions, it categorizes memory into three layers: an explicit persona layer (to replace default presets) and a separate contextual section for character information and general memory. Compatible with DSH version 0.1.1 and licensed under the MIT License, its core features include reading and writing memory via the `get_session_memory` and `update_session_memory` interfaces, Chinese rendering of model fields, lossless migration of V1-V3 data to the V4 sidecar format, and default settings.
Read Moremindspeace-dsh-local-rag: Local Hybrid RAG Plugin for DeepSeek Harness
mindspeace-dsh-local-rag is a local hybrid RAG plugin designed for DeepSeek Harness, licensed under the MIT License. It registers the `search_local_memory` tool, allowing the model to autonomously decide to retrieve local materials when context is insufficient, avoiding the injection of irrelevant context or reliance on remote APIs in every round. The plugin combines vector similarity with BM25+ lexical retrieval and supports a purely lexical fallback when vectors are unavailable. It manages two types of corpora: user-uploaded multi-format files (such as PDF and DOCX) and session compressed summaries, all of which support viewing, editing, and version rollback. The plugin features built-in local model lifecycle management, supporting the download, verification, and persistent startup of approximately 407MB ONNX models, with internet access limited to the model download phase. Installation requires...
Read Moredsh-worktime-board: DSH Local Work Hours Board and Cultivation Realm Progression
dsh-worktime-board is an MIT-licensed DSH plugin maintained by spacexun2, aimed at solving the inconvenience of aggregating Agent runtime metrics across threads and days. By recording active duration, model tokens, user input, and tool calls, it transforms workload into "Cultivation Value," presented in stages ranging from Qi Refining to the Twelve Realms of the Primordial Chaos. The plugin offers various views including daily, weekly, monthly, and academic year (August 1st to July 31st of the following year), supports viewing 24-hour distribution, the Top Performers List, and thread attendance, and allows users to navigate to corresponding sessions. Furthermore, it registers a `worktime_summary` tool for Agent queries and provides an HTTP interface to retrieve status data. Upon installation, data is stored locally in `$DSH_HOME` using 5-minute slot counting.
Read Moredsh-memoryhub: Integrate MemoryHub project memory into DeepSeek Harness
dsh-memoryhub is an MIT-licensed plugin for DeepSeek Harness (dsh), designed to address the issue of project memory usage across sessions. It integrates the `.memoryhub/` checkpoint memory maintained by MemoryHub into dsh sessions, supporting bidirectional data synchronization: automatically loading checkpoint memory upon session startup, while also allowing the current session to be saved back to MemoryHub via the `mh_save` command. The plugin registers six core tools (such as `mh_load`, `mh_save`, `mh_status`, etc.) and the `mh` workflow skill, providing a Memory tab in the Web UI for visualizing checkpoints.
Read MoreDSH Plugin: Conversation Indicator, Marks User Messages Next to the Web GUI Scrollbar
`@smanx/dsh-conversation-indicator` is a community-driven third-party plugin for the DeepSeek Harness (DSH) Web GUI, designed to address the pain point of quickly locating user messages in long conversations. The plugin displays a compact, capsule-shaped indicator panel near the scrollbar on the right side of the chat area, marking only user input lines. Users can hover over it to view a summary, and clicking it triggers a smooth scroll to the corresponding location. The panel automatically hides during scrolling or streaming output, appears upon mouse proximity, and fades out upon mouse leave to avoid obstructing normal interactions. As a pure client-side plugin, it does not alter model behavior, incur additional token usage, modify tool calls, or record session logs; preference settings are stored in the browser's `localStorage`. Installation requires adding the GitHub repository via the `dsh plugin` command.
Read MoreDeepSeek-harness-lark: Integrating Feishu and Lark text, image, and file channels for DeepSeek Harness
sliverp/DeepSeek-harness-lark is a DeepSeek Harness channel plugin based on the MIT License, designed to connect Feishu and international Lark to the DSH ecosystem via official WebSocket transmission without the need for a public callback server. The plugin supports handling text, rich text, images, and standard files, and maintains persistent and isolated sessions. Core features include automatic reconnection WebSocket transmission, support for downloading files to the workspace and sending files during replies, and the integration of Harness slash commands (such as `/new` and `/approve`) to enable session management, tool approval, and status checks. Configuration requires Node.js 22.19+ and DSH 0.1.0-rc.7+, and must be set up within Feishu.
Read Moredeepseek-harness-dingtalk: DingTalk Stream Channel Plugin for DeepSeek Harness
`deepseek-harness-dingtalk` is an MIT-licensed DingTalk channel plugin designed to integrate DeepSeek Harness (DSH) agents with DingTalk. Leveraging the official Stream WebSocket connection, this plugin eliminates the need for a public network callback server and requires only the configuration of a Client ID and Secret to establish the connection. In terms of core capabilities, the plugin supports receiving text, Markdown, rich text, speech-to-text, images, and standard files, while securely handling media downloads and multimodal inputs via the official API. On the output side, it supports DingTalk native Markdown replies, images, and workspace file uploads, and can fully execute the Harness Agent Loop. The plugin provides in-session tool approval (via one-time codes),
Read Moredsh-enhanced-plugins: Optional extension functions for DSH
dsh-enhanced-plugins is an enhancement suite for DeepSeek Harness (DSH) designed to add optional features to DSH via public plugin extension points without modifying core code. The suite consists of seven independent Cordis feature packages and a Windows Launcher, supporting full installation or selective installation based on stable Feature IDs. Key features include Windows desktop entry, sound alerts and desktop pets, community plugin management, MCP server management, model input type declaration, last message editing, sub-agent toggles, and official Team monitoring. Installation requires Node.js 22.19+ and DSH 0.1.2-alpha.2, while Windows-specific features require Windows 10.
Read Moredsh-comfyui-node-development: Packaging the ComfyUI custom node development guide into a DSH skill
dsh-comfyui-node-development is a DeepSeek Harness (DSH) plugin designed to provide a standardized reference for ComfyUI custom node development for DSH agents. This plugin condenses the official ComfyUI documentation into a loadable runtime skill, comfyui-node-dev, covering core topics such as node class contracts, input/output types (INPUT_TYPES/RETURN_TYPES), frontend JavaScript extensions, internationalization, data, and V1 to V3 migration. Through built-in reference manuals like backend.md and frontend.md, as well as bilingual summary documents, agents can automatically load relevant materials on demand during conversations, resolving context pollution and redundancy caused by temporary consultations.
Read Moredsh-plugin-multimodal: Adding image support to text-only DSH sessions
dsh-plugin-multimodal is a visual sidecar plugin for DeepSeek Harness (DSH), designed to address the issue where text-only main models cannot process image inputs. This plugin receives image attachments via the GUI, invokes a vision-capable sidecar model to convert image content into text descriptions, and then sends the text to the main model, thereby bypassing the "current model does not support images" limitation. Core features include support for pasting images in the Web UI, a see_image tool for processing disk screenshots, and configuration of the vision model via environment variables or configuration files. Users need to install the plugin and restart the DSH Web process to ensure the sidecar points to a real multimodal model (such as GLM-4.5V) and avoids using a text-only model as the vision endpoint. This plugin is suitable for those who wish to...
Read More@shatyuka/dsh-llm-codebuddy: Integrate Tencent CodeBuddy login and model list for dsh
`@shatyuka/dsh-llm-codebuddy` is a DeepSeek Harness (dsh) plugin maintained by shatyuka and licensed under the MIT License, designed to seamlessly integrate Tencent CodeBuddy into the dsh ecosystem. Its core advantage is that it requires no API key, allowing users to access the CodeBuddy model list by logging in directly through a browser. The plugin supports automated credential injection via the `Sign in` button in the Web UI's `Settings → CodeBuddy`, as well as login, status checks, and logout operations via the terminal using the `dsh-codebuddy-login` command. Ensure your local Node version is `^22.1` before installation.
Read Moredsh-utility-tools: Adding file drag-and-drop and text referencing to DSH chat
dsh-utility-tools is an input-side utility plugin for DSH conversations, maintained by sharkymew and licensed under the MIT license. The plugin primarily addresses two issues: first, it supports dragging and dropping arbitrary files into the chat area, allowing for the preview of filenames and sizes before sending, as well as individual removal; second, it allows users to select conversation text and add it to the beginning of the pending message as a blockquote. The plugin provides two installation methods: static mounting and dynamic loading. The static version installs to the web profile via a single command and utilizes HTTP routing for uploading attachments; the dynamic version loads with zero dependencies using `cordis_define` and runs in-process. It is applicable for scenarios in web or Electron environments where file attachments or context references need to be added. Please note that standard browsers cannot access real disk paths, and the single file limit is 8MB.
Read Moredsh-yolo-mode: Add LLM-based automatic approval for DSH sandbox privilege escalation requests
dsh-yolo-mode is an MIT-licensed plugin for DeepSeek Harness (DSH) designed to address the issue where frequent privilege escalation approvals disrupt workflows within sandbox sessions. When the approval strategy is `ask` and the system is in specified sandbox modes (e.g., `workspace-write`), this plugin leverages Large Language Models (LLMs) to automatically adjudicate privilege escalation requests such as `danger-full-access`, thereby replacing manual confirmation steps. It offers presets including `off`, `strict`, `balanced` (default), `permissive`, and `yolo`, and supports configuring custom permission levels via `levels` to set allow, deny, or human-escalation strategies for different target modes or specific tools (e.g., `pwsh`, `write`).
Read MoreSecurStack DSH Plugin: Invoking SecurStack Scanning, Policies, and Diagnostics in DeepSeek Harness
The SecurStack DSH plugin is an adapter for DeepSeek Harness, designed to enable agent workflows to directly invoke the existing `securstack` CLI without the need to re-implement security tools. The plugin primarily provides four core functions: executing repository security scans via CLI, performing policy threshold checks, running local diagnostics, and returning structured JSON results when the CLI supports it. It reuses existing SecurStack authentication methods (such as API Key or command-line login) and strictly adheres to conservative boundaries; version 1 does not include low-level logic such as scanning engines, encryption, or Shielding operations. Ensure the environment is Node.js 20 or higher before installation and enable it via the `dsh plugin` command. The plugin is suitable for...
Read Moredsh-soul-md: Persona Cards and Long-term Memory for DeepSeek Harness
dsh-soul-md is a community plugin for DeepSeek Harness (DSH) designed to centralize the management of personas and long-term memory via a settings interface, avoiding the manual maintenance of scattered files. Its core functions include "Persona Cards" and "Long-term Memory": Persona Card content is rendered into the system prompt, supports multiple cards, switching by conversation or workspace, and allows the AI to read, write, and evolve via tools. Long-term memory provides persistent storage, is subject to size limits when injected into prompts, and prioritizes association with the currently active persona card. Installation requires Node.js >=20.18, and the new version requires DSH >=0.1.0-rc.7; users with older versions need to lock to specific versions. In terms of usage, users can create cards and specify default values on the settings page, while the AI autonomously manages memory and personas using dedicated tools. This plugin is suitable for scenarios requiring...
Read Moredsh-plugin-marketplace: Browse community plugins in the DSH Web UI Settings page
`dsh-plugin-marketplace` is a DSH Web UI plugin maintained by Scorp1o117, designed to streamline the tedious process of frequently switching between GitHub and the terminal to find and install DSH plugins. By leveraging the public GitHub Search API, this plugin integrates a plugin marketplace directly into the DeepSeek Harness Web UI settings page, supporting keyword searches on `github.com/topics/dsh-plugin` with options to sort by star count or update time. Users can directly browse plugin cards (containing descriptions, languages, update dates, etc.) and detail panels (containing README summaries, installation commands, and repository links), and even utilize the "AI explain" feature via a configured large model...
Read More