dsh-agent-contract：多智能体契约工作流引擎
dsh-agent-contract 是 DSH 工作流类插件，由 t2094308-star 维护。它在派活时为子代理现场装配契约提示词（总纲、角色卡、任务与文档切片、工具白名单），主代理只收短回执；交作业自动补文档头部并记账，审计红黄清单只提醒不拦截。支持图书管理员治理、侧栏契约面板与项目目录自适应。本地安装需 npm install 后 dsh plugin link，完整退出应用后生效。
Read Moredsh-team-employees：把 DeepSeek Harness 变成多员工协同团队
dsh-team-employees（0.5.0，MIT）由 syc67169-source 维护，把工作流插件装进 DeepSeek Harness：五位员工 preset、$DSH_HOME/team 花名册与纯文本账本、按阶段推进的内容流水线及仅人工可用的 content_gate，侧边栏办公室反映真实会话工位状态；0.5.0 起支持本地长期研究职责。安装需 Node 20+，可通过 link-into-profile 脚本或手动改 profile 的 github 依赖与 bundles，完全重启 DSH 后验收。适合需要人工审核与发布确认、且不期望插件代发平台的团队。
Read More帝国架构 Empire Architecture：用「三公九卿」组织多智能体协作
帝国架构 Empire Architecture 是 SkillHub 社区 ai-agent 类 Skill（v1.0.0，@user_168d274d/empire-architecture-v2），以中国古代三公九卿隐喻组织多智能体，简介涵盖 256 节点、标签路由、模型分级、任务队列、Agent 记忆与量子计算思维模拟器。本文依据 SkillHub 元数据与安装文档说明 skillhub CLI 安装、Cursor/Claude Code/Codex 的 skills 目录，并注明 requires_api_key 与社区来源注意事项；公开页面未提供独立可复现示例，需安装后阅读 SKILL.md。
Read Moredsh-plugin-roundtable：把 DSH 会话做成可视化专家圆桌
dsh-plugin-roundtable（RoundTable）是 DeepSeek Harness 的工作流插件，由 Fishsb 维护，需 DSH 0.1.5-rc.1+，MIT 许可。它将单会话组织为主持人、多席子代理专家与汇聚网关的圆桌会议，Web UI 提供拓扑与群聊视图，支持主持人统筹、实验性多模型平等与针锋相对红队三种模式，以及调度面、人类决策卡片与 Markdown 导出。安装需从源码 pnpm build 后 dsh plugin add。附带 28 席 experts.yaml 可选导入。使用前请审阅源码；插件以 dsh 进程权限运行。
Read Moredsh-devflow：在 DSH 会话里跑多智能体交付工作流
DevFlow（@xiaoxie-ide/dsh-devflow）是 BinXie-lixiangdehua 为 DeepSeek Harness 开发的工作流插件：Commander 与固定/临时角色、dispatch→execution→audit→rework→acceptance 状态机、devflow_* 调度工具（并发上限 5）及右侧可视化面板；状态持久化在工作区 .devflow。需 Node ^22.19.0 或 >=24、dsh >=0.1.7-rc.1。Windows 下可用官方 PowerShell 一键 install.cmd 安装并选用 DevFlow preset。MIT 许可。
Read Moredsh-office：在 DeepSeek Harness 里搭一间持久的「同事办公室」
dsh-office 是 windwhiterain 为 DeepSeek Harness 开发的第三方工作流插件（0.1.0，MIT）。它将普通 DSH 会话变为带花名册的「同事」，办公室层持久保存频道、私信与用户 mailbox，并通过 wake 与 notify 在合适时机投递或唤醒。安装使用 README 中的 plugin_manager install_bundle target: github:windwhiterain/dsh-office。本文基于官方 README 介绍招聘、office_post/office_dm、Web Office 面板及适用场景，并提醒安装前审阅源码与进程权限。
Read Moredsh-agents：23 个专职角色与三档模型路由的 DSH 多智能体工作台
dsh-agents 是 wangyuanchuan2022 维护的 DSH 社区插件，通过 agent_roles、agent_taskbook、agent_spawn 将 23 个专职角色与 LOW/MEDIUM/HIGH 档位路由、八节任务书合成和标准子会话创建串成一次调用。含 25 个工作流技能、评审隔离与机检工具。本文基于官方 README 介绍环境要求、安装命令与快速开始示例，并提醒插件以 DSH 进程权限运行、安装前需审阅源码。
Read MoreDSH expert-team 插件：把多角色专家团装进当前会话
expert-team 是 kee0012 维护的 DSH 工作流插件，提供 25 支内置专家团与团队画廊，一键召唤后在独立子会话中并行工作，主会话作为召集人汇总交付；支持流水线阶段派遣与本地自定义 teams.json。本文基于 GitHub README 介绍安装命令、界面召唤步骤、斜杠命令、模型工具 expert_team_summon 及配置与已知限制。
Read Moreagentscope-builder：用 Agent Skill 按官方 API 搭建 AgentScope 2.x 应用
agentscope-builder 是 SkillHub 社区 Skill（@user_ea778c49/agentscope-builder，v1.0.2），面向 AgentScope 框架搭建指南：默认 2.x，含版本核对、分主题 reference、最小单智能体模板与本地校验脚本。文章说明其解决 1.x/2.x API 混用痛点，给出 skillhub 安装命令、Cursor 等 skills 目录，以及最小工程与环境检查的可复现步骤；并注明 requires_api_key 与示例基线 AgentScope 2.0.10。
Read More涌现智能编程：DeepSeek TUI 多智能体自主开发助手——在 Agent 里委派 DeepSeek 写代码
SkillHub 社区 Skill @user_1acc1bde/emergence-deepseek-tui（v1.0.0，dev-programming）教 Agent 用 DeepSeek TUI CLI 做自主编程：exec 模式无头纯文本委派，由主 Agent 提供上下文并自行落地；run 模式开启完整工具与文件系统访问。文章说明安装 skillhub 与 deepseek-tui、Cursor 等 skills 目录、典型命令与 exec/run 选型注意点；平台标注 requires_api_key=true。
Read More高效文档处理多智能体团队：用五个 Agent 串起文档读写与报告
SkillHub 社区 Skill「高效文档处理多智能体团队」（@user_d76c672d/doc-processing-team，v1.0.3）基于 agent-team-template v2.1.0，由 5 个专业 Agent 协作完成 PDF/Word/Excel/图片的读取、英译中、分类、总结与报告生成。覆盖 PDF 转 Word/Markdown、批量归档、合同审阅、OCR、文献整理、文档问答与信息抽取等场景；声明国内可离线、不上传敏感数据，且 requires_api_key=false。本文依据 SkillHub 官方简介与安装文档说明 skillhub 安装方式及 Cursor、Claude Code、Codex 等 skills 目录要点。
Read MoreDeepSeek Harness 圆桌讨论插件
DeepSeek Harness (DSH) 的 `dsh-roundtable` 插件提供了一套多智能体圆桌讨论机制，旨在降低多智能体协同编排成本。该插件支持真实 Subagent、卡片式引导、普通聊天流输出及自动化纪要生成。用户通过侧边栏发起讨论，按序添加成员角色和模型，插件将逐个启动成员发言，并由主持人汇总，最终输出结构化 Markdown 纪要。安装后需重启 DSH Desktop。
Read Morerolebox：将单一编码智能体配置为工程团队
`rolebox` 是一个适用于 DeepSeek Harness 等工具的通用多智能体插件，通过 YAML 定义与 Graph v3 协议将单一 Agent 转化为具备独立提示词、技能与权限的工程团队。该插件支持 18 个预设角色、持久化 SQLite 记忆以及 32 个 LSP 和桌面控制工具。核心功能基于声明式工作流（`graph_declare` 与 `graph_submit_outcome`）实现多角色协作与循环任务管理。安装需 npm 包，支持在 DSH 设置中查看角色。插件适用于需要严格工作流控制与角色分工的复杂工程任务，但需注意权限风险及声明式执行的限制。
Read More多智能体小说创作系统：把长篇创作拆成可验收的六步流水线
介绍 SkillHub 社区技能「多智能体小说创作系统」（@user_6d9676ae/novel-multi-agent-skill，v1.0.0）：以总策划、世界观、角色、大纲、写作、评审多角色分工，配合 templates、rules 与 workflow 脚本形成可验收的闭环。说明 SkillHub CLI 安装到 Cursor/Codex/Claude 的 skills 目录，以及 OpenClaw 下 init-project 与 step 脚本典型用法，并提示 agents 需自行配置、性能指标以官方文档为准。
Read Moredsh-plugin-devin: Devin CLI Proxy Tool in DeepSeek Harness
`dsh-plugin-devin` is an unofficial bridge tool designed to integrate Devin CLI capabilities into DeepSeek Harness (DSH). It exposes Devin’s headless print mode as a DSH tool via `ctx.subprocess`, and supports LLM provider routing, image processing, session management, and timeout control. Users must install the plugin and configure parameters in a Profile to connect Devin as a sub-agent, making it suitable for scenarios that require delegating complex tasks to Devin.
Read MoreDeepSeek Harness Expert Orchestration Plugin
`dsh-expert-orchestrator` is a multi-agent orchestration plugin for DeepSeek Harness (DSH) designed to address task confusion and lack of quality control in complex collaboration. It enforces strict workflow management through PM-first planning, a dependency DAG task board, and staged delivery gates. Core features include per-round triage, expert delegation with a merge roster, parallel processing, and an experience pool mechanism. In addition, the plugin provides delegation safety gates, crash-safe state, and tool budget recovery, making it suitable for complex development scenarios that require high-quality delivery.
Read Moredsh-agent-teams-fish
`dsh-agent-teams-fish` is a visually enhanced supplementary version of `NanmiCoder/dsh-agent-teams`, designed to address the "face" problem in multi-agent collaboration scenarios. While preserving the original natural-language team formation, DAG task scheduling, and mailbox messaging logic, the plugin significantly enhances its artistic presentation. It ships with 286 built-in avatar assets from 17 vendors and 10 job roles, supports routing by vendor and role, and provides a custom art directory feature to flexibly adjust the visual style. Additionally, the plugin has specific version requirements for the DSH environment and includes built-in trademark notes and security mechanisms.
Read Moremeow-dsh-workflow: DSH Plugin: Building Standardized Collaboration Pipelines with a Role Library and Sub-Agents
`meow-dsh-workflow` is a DeepSeek Harness (DSH) plugin designed to reduce the burden of agent prompt maintenance and permission management. It uses a built-in editable role library to compile role definitions into system prompts in real time, and leverages `meow_agent_call` to apply hard permission filtering to sub-agents. Core features include cross-session role memory, a company ledger, Web dashboard management, and skill catalog filtering. The `web` profile must be specified during installation. It is recommended to use it together with `meow-dsh-task`: the former tracks “tasks,” while the latter coordinates “people,” enabling a structured multi-agent workflow.
Read MoreDeepSeek Harness Multi-Agent Group Chat Plugin: dsh-groupchat
`dsh-groupchat` is a DeepSeek Harness plugin designed to address multi-agent collaboration and task management. It allows users to create independent group chats within a single interface, supporting cross-group collaboration, customizable AI members, task allocation, and memory sharing. By configuring a global font size and task board, the plugin optimizes the multi-role collaboration experience. Users must install it via a local link and enable it in the host. The plugin is not affiliated with the official project and relies on the host service to access the API.
Read Moredsh-ultracode: A Multi-Agent Workflow Orchestration Plugin for DeepSeek Harness
The `dsh-ultracode` plugin for DeepSeek Harness (DSH) uses a plugin-based architecture to inject authorization banners and instructions, providing Composer with three modes—`off`, `high`, and `ultra`—for explicit control over multi-agent workflow orchestration during a session. When enabled, the plugin requires the model to keep outputs concise and return materials as files, while also supporting instruction presentation and reasoning protection. Users can install the plugin via npm and switch between modes using the `/ultracode` command, while taking note of Node.js version and permission/security requirements.
Read Moredsh-eteams: DeepSeek Harness Multi-Agent Team Collaboration Plugin
The ETeams plugin for DeepSeek Harness (DSH) uses a multi-agent architecture and improves workflow efficiency through a model in which a lead breaks down objectives and members execute them in sequence. It supports custom roles, differentiated model configurations, and multi-session collaboration, with all state data stored in a local SQLite database to ensure privacy and security. Users can install it via the command line and use the /eteam panel to assemble teams and track task progress, addressing the pain points of objective breakdown and allocation in complex task processing.
Read Moreoh-my-deepseekharness
`oh-my-deepseekharness` is a multi-agent orchestration plugin compatible with DeepSeek Harness. It encapsulates the OMC/OMX orchestration concept and provides three mutually exclusive execution modes—`autopilot`, `ralph`, and `team`—supporting end-to-end collaboration, from natural-language automated orchestration to explicit iterative pipelines. The plugin integrates 33 role cards, 42 skills, and 55 MCP tools, and features a DAG runtime, code intelligence (AST, LSP), and cross-session memory. Installation requires Node.js 22.19.0 and dsh 0.1.7+, is done via npm, supports flexible configuration, and recommends ignoring the `.omd/` directory in version control.
Read MoreDeepSeek Harness Plugin: Multi-Agent Group Collaboration
The DeepSeek Harness plugin `polohot/dsh-adrian-agents-group-work` provides DSH with multi-agent collaboration capabilities. It converts a single session into a room comprising a Chair (lead) and Members (sub-sessions), supporting member-to-member messaging, task management, and JSON-based communication. After installation, users can start a room using natural language or slash commands (such as `/agents-group-work`), where the Chair parses the task and coordinates members to complete the work. The plugin features state persistence, progress monitoring, and safety constraints (such as single-room support only and member naming conventions), making it suitable for scenarios requiring coordinated agent collaboration.
Read Moreoh-my-dsh: DeepSeek Harness Multi-Agent Orchestration Layer
`oh-my-dsh` is a multi-agent orchestration plugin based on DeepSeek Harness (DSH), designed to replicate proven workflows. By encapsulating 30 auto-loaded skills and 13 slash commands, it enables zero-configuration team-first orchestration, automatic parallelization, and natural language interaction. Key features include inline HUD progress visualization and experience-driven learning. Built on DSH primitives with no additional runtime overhead, it is well suited for rapidly building multi-agent collaboration systems.
Read MoreETeams: DeepSeek Harness Multi-Agent Team Collaboration Plugin
The ETeams plugin for DeepSeek Harness adopts an "everything is a plugin" architecture to overcome the logical reasoning limitations of a single model. Developed by developer 2507483326, it enables multi-agent team operation through real multi-session collaboration rather than simulation. The plugin supports forming teams with a leader and members, and uses independent sessions, task slots, and execution chains to break down and execute complex tasks. It features self-contained builds, local SQLite persistence, and dual-end support, ensuring data privacy and operational convenience. Users can create roles with the `/eteam` command, form teams in the panel, and enable automatic task assignment and real-time progress monitoring.
Read MoreIntroduction to the DeepSeek Harness (DSH) Plugin: dsh-intercom
`dsh-intercom` is a plugin for DeepSeek Harness (DSH) that addresses communication and coordination among multiple DSH process sessions on the same machine. It is maintained by Sqhao-O and released under the MIT license. The plugin provides session discovery, 1:1 messaging (with send, ask, and reply support), alias configuration, and a Web UI panel. All sessions communicate through a local proxy process; after installation, the `intercom` tool can be used to enable cross-process interaction, making it suitable for scenarios requiring coordination of multiple sessions.
Read Moredsh-teams-x: Transforming Single Sessions into Coordinated Multi-Agent Teams
dsh-teams-x is a multi-agent team plugin for DeepSeek Harness, designed to handle complex tasks by coordinating multiple agents. The plugin decomposes high-level goals into dependency-aware task DAGs and supports persistent mailbox communication, automatic repair loops, and quality-gate contracts. Core features include a Web-based planning editor, an all-SVG activity panel, team templates, and in-session card display. Installation requires restarting the host, the tool namespace is `teamsx_*`, and it natively supports DSH 0.2.0-rc.1 or later.
Read Moredsh-agent-teams: Turning DeepSeek Harness Sessions into Multi-Agent Collaborative Teams
dsh-agent-teams is a DeepSeek Harness plugin that allows the current session to act as a captain and orchestrate multiple persistent sub-agent members, achieving complex task decomposition, dependency management, and result aggregation through natural language. It features captain delegation, dependency-aware tasks, direct communication, quality gating, and a real-time activity panel, enabling multi-agent collaboration within a single session without the need for an additional workflow engine. It is recommended to run in the Harness 0.1.5-rc.1 environment.
Read Moredsh-a2a: Bidirectional Agent2Agent v1.0.1 Plugin for DeepSeek Harness
`dsh-a2a` is a DeepSeek Harness plugin that transforms DSH into a multi-face gateway supporting the A2A v1.0.1 protocol. It provides multiple inbound servers (with auto-derived skills), multiple outbound connections, and full-lifecycle management through a GUI. It supports persistent task storage, multiple execution modes, and authentication management, creating a complete closed loop from protocol implementation to operations and maintenance, and is suitable for multi-agent scenarios requiring bidirectional A2A interaction.
Read MoreDeepSeek Harness (DSH) Plugin: dsh-agent-frugality
`dsh-agent-frugality` is a DeepSeek Harness (DSH) plugin designed to address excessive token consumption and limited observability in multi-agent systems. Through read-ledger deduplication, compression-rule immunity, mandatory completion gates, and low-cost review mechanisms, it constrains agent behavior and controls costs. Installation requires Node.js 20 and DSH 0.1.0-rc.6, and the plugin supports flexible configuration via tools and environment variables, enabling effective quantification of waste from repeated reads and prevention of rule loss.
Read Moredsh-team-superflow: Fixed Five-Person Development Team and Spec-Superflow State Machine
`dsh-team-superflow` is a multi-agent collaboration plugin for DeepSeek Harness. Rebuilt from `spec-superflow`, it provides five fixed members (spec writer, architect, engineer, reviewer, and documentation maintainer) and an eight-stage standard workflow. The plugin implements automatic stage progression, build gates, and visualized web tracking, supports creating tasks, managing members, and rolling back processes through commands, and aims to simplify the automated development process from requirements to delivery.
Read MoreDeepSeek Harness Plugin: dsh-subagent-progress Real-Time Subagent Progress Dashboard
`dsh-subagent-progress` is a client-side plugin for DeepSeek Harness, designed to address the challenge of real-time visibility into subagent status during multi-agent development. It aggregates subagent session events into a live dashboard, with cards displaying progress, elapsed time, and specific actions. The plugin supports subagent self-reporting, provides alerts for suspended or failed states, and offers intervention controls such as retry and pause. The card layout is clean, supports hover to view details, and includes automatic cleanup. Installation is straightforward and requires no configuration, significantly reducing the overhead for developers to monitor multi-agent runs.
Read Moredsh-agent-hub: Group-Chat-Style Multi-Agent Collaboration Console
`dsh-agent-hub` is a plugin for DeepSeek Harness that provides a group-chat-style collaborative console for multi-agent coordination. It transforms the collaboration process into a message stream, featuring a central orchestrating brain, atomic task decomposition, targeted dispatch, a validation-and-rework closed loop, and autonomous optimization with circuit-breaking mechanisms. Users can dispatch tasks by `@mentioning` members, while the brain autonomously advances tasks through to completion and knowledge consolidation. The plugin supports zero-dependency deployment, offers full visibility and traceability, and enables multi-round iteration with human intervention.
Read MoreIntroduction to the DeepSeek Harness MoA Plugin
`dsh-moa` is a multi-agent collaboration plugin for DeepSeek Harness (DSH), designed to address the limitations of single models in complex tasks through a Mixture of Agents (MoA) architecture. It supports the `/moa` command and offers core features such as adaptive question gating, parallel proposals, review mechanisms, disk-level candidate isolation, and incremental editing. It also provides native settings cards, quick mode, and win-rate leaderboards, and can integrate with Live Canvas. By promoting the best solution among multiple models to the project root directory, the plugin significantly improves the reliability and quality of code generation.
Read Morecaptain-dsh-plugin: Multi-Agent Professional Navigation Workflow on DeepSeek Harness
captain-dsh-plugin is a resumable, multi-agent workflow plugin for DeepSeek Harness designed for recruitment screening. It transforms ambiguous decisions into traceable reports through Agent collaboration and evidence tracking. The plugin features core capabilities including Agent orchestration, context engineering, checkpoint recovery, and strict security and access control. It supports real-time control, automated data entry, and version management, while strictly adhering to principles of privacy protection and factual boundaries to ensure a reproducible and secure workflow.
Read Moredsh-cross-session-agent: DeepSeek Harness Cross-Session Agent Coordination Plugin
`dsh-cross-session-agent` is an unofficial plugin for DeepSeek Harness designed to address conflicts and coordination issues in parallel multi-agent development. Through peer discovery, message passing, transport receipts, and restricted context projection, it enables agents to securely negotiate file ownership, report progress, and request handoffs. Its core capabilities include discovering peers, sending coordination directives, querying transport status, and reading restricted contexts. Strict permission boundaries must be maintained during use: only regular Sessions in the same workspace as the current one may be read, and reading sensitive reasoning data or executing commands contained in messages is prohibited. The plugin provides the infrastructure for multi-agent collaboration through its safety mechanisms.
Read Moreevo-subagent: DeepSeek Harness Plugin: Role-Based Subagent Routing and Evolution
The `evo-subagent` plugin for DeepSeek Harness (DSH) is designed to address pain points in routing and experience accumulation in multi-agent systems. It supports creating and managing role-based subagents, using an evolution mechanism to maintain `prefercmd.md` and `memory.md`, thereby accumulating reusable commands and experience while saving tokens. The plugin supports binding and routing, strict validation, workspace isolation, and `spawn`/`fork` execution modes, and provides security isolation mechanisms. Installation requires DSH `0.1.5-rc.3` and Node.js `>=24.2.0`. Users can configure Markdown binding files to define roles, utilize built-in roles (such as the code reviewer) and `/evo-route-doctor`, among others.
Read Moredsh-plugin-agent-dashboard: DeepSeek Harness Multi-Agent Status Dashboard
dsh-plugin-agent-dashboard is a multi-agent dashboard plugin for DeepSeek Harness, designed to address the challenge of monitoring multi-agent parallel execution. It allows agents to output structured status reports through instructions, including runtime status, task dependencies, progress, and time information, without requiring a direct web interface. Maintained by YYTbit and released under the MIT license, the plugin can be installed via plugin management commands and is suitable for scenarios that require real-time monitoring and reporting of agent progress.
Read Moredsh-agent-teams: Transforming DSH Conversations into Multi-Agent Teams
`dsh-agent-teams` is an open-source plugin designed to enable DeepSeek Harness (DSH) sessions to act as team leaders automatically through natural language instructions, coordinating multi-agent teams to complete complex tasks. It provides core capabilities such as leader delegation, persistent members, dependency-aware task management, direct messaging, and a real-time Web UI. Its single-leader mechanism and file-based state storage limit concurrent processing, but it offers a lightweight, structured collaboration solution for building multi-agent systems within DSH.
Read Moredsh-shadow-mind: A Parallel Cognition Runtime for DeepSeek Harness
`dsh-shadow-mind` is a plugin for DeepSeek Harness (DSH) designed to provide the main agent with parallel cognitive capabilities. It uses a heartbeat scheduler to randomly activate configured “shadow agents” after a main-agent turn. These agents have independent responsibilities, tool permissions, and timeout limits, and can assist the main agent with review or fact-checking. The plugin supports restricted tools, lifecycle management, automatic pause/resume, and sensitive-parameter masking. It is easy to install and well suited for complex workflow scenarios that require multi-perspective assistance.
Read MoreEvolving-Werewolf: An AI Werewolf Engine with Cross-Game Evolution
Evolving-Werewolf is a DeepSeek Harness plugin maintained by VegeFin, offering a complete 9-player Werewolf engine solution. Its core feature is cross-game knowledge accumulation: after each game, the AI reviews the match, records the experience in a knowledge base, and improves its strategy in the next game through persona injection, while relying on a state machine to manage the game flow and prevent LLM information distortion. The plugin supports full character skills, a standard daytime process, information isolation, and anti-freeze mechanisms, and integrates Zhipu AI for avatar generation. Through the `werewolf_*` tools and a browser panel, users can control the game progress or participate as human players, making it suitable for simulating multi-player logical confrontation in DSH.
Read MoreDeepSeek Harness Plugin: dsh-agentmonitor Multi-Agent Monitoring Dashboard
dsh-agentmonitor is a multi-agent monitoring plugin for DeepSeek Harness (DSH). It provides a visual monitoring console via a sidebar, featuring a dual-half structural design, a self-hosted static dashboard, and zero external dependencies. The plugin must be installed using a specific command and displays sample data by default. Users can connect real-time monitoring data by configuring `dashboard-data.js` and `dataDir`, making it suitable for observing agent runtime status in real time within a DSH environment.
Read Moredsh-jj-vcs: Jujutsu Version Control Plugin for DeepSeek Harness
The DeepSeek Harness (DSH) plugin `dsh-jj-vcs` addresses version-control bottlenecks in multi-agent collaboration by introducing the Jujutsu version control system. It provides each agent with an isolated workspace and leverages a change-based model to deliver clear attribution auditing, concurrency safety, and operation tracing. Its core features include creating independent changes, explicit permission tiers, controlled publishing, and Windows compatibility. Developers must install it manually and depend on JJ 0.44.0, using the coordinator script to start tasks and verify changes, thereby providing a robust version-management solution at the community level for enterprise-grade multi-agent projects.
Read MoreDeepSeek Harness Plugin: ttxl314/dsh-skill-lord-serf
ttxl314/dsh-skill-lord-serf is a plugin for DeepSeek Harness (DSH), designed to address challenges in multi-agent collaboration such as state persistence, protocol inconsistency, and difficult validation. The plugin introduces a file-based multi-agent orchestration protocol, provides dual-role skills—Lord (orchestrator) and Serf (executor)—and leverages the file system for state storage and Schema validation to ensure process traceability. It supports the full lifecycle from planning to decision-making and enables reliable distributed task execution through standardized task-package distribution and reporting mechanisms.
Read MoreTaskSwarm: Multi-agent Task Orchestration Plugin for DeepSeek Harness
TaskSwarm is a multi-agent task orchestration plugin for DeepSeek Harness (DSH) designed to automate the handling of dependent tasks. Its core mechanism involves stratifying tasks into Waves based on dependencies and executing them in parallel via independent Lanes within the same layer. Each Lane is isolated using Git Worktree and merged into the `taskswarm/orch` integration branch, supporting crash recovery and semantic conflict resolution. The plugin provides Supervisor coordination, independent Reviewer reviews, and LLM Merger merging functionality, while achieving asynchronous communication via a file mailbox. It features a built-in local Web Dashboard based on SSE to display status in real-time. Installation requires adding the npm package via the `dsh plugin` command.
Read Moredsh-my-go: DSH-based Multi-agent Orchestration Plugin for Sisyphus
dsh-my-go is an agent orchestration system built on top of DeepSeek Harness (DSH) and adopts the MIT license. The system unifies the management of 7 professional sub-agents (such as Hermes and Oracle) through the Sisyphus scheduler, adopting a star topology structure to ensure that only one sub-agent runs at any given time. Its core capabilities include flexibly configuring the provider, model, and reasoningEffort based on job types, supporting off-peak routing based on Beijing Time to optimize resource usage, and providing communication tools such as go_work as well as the Sisyphus quality control mechanism. Model configurations are verified before dispatch, and if they do not exist, they fall back to parent session routing. The system provides a WebUI settings page and a tree diagram panel for convenient visual management. Installation requires...
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-virtual-product-team: Driving DSH's Product Pipeline Through Dialogue
`songoao25/dsh-virtual-product-team` is a conversational mode preset for DeepSeek Harness (DSH), designed to automate the full product lifecycle management through a multi-agent workflow. The tool decomposes tasks into 12 stages across 8 phases, including idea validation, requirement definition, technical design, development QA, release, marketing, and iteration. Its core mechanism employs "stage gating," where progress is reported to the user at the end of each phase, and advancement to the next stage occurs only after user approval. Users are only responsible for decision-making and approval, requiring neither technical expertise nor coding skills. For installation, it supports a full installation via `./install.sh` or installing only the skill pack under a specified profile. When in use, users need to start a new conversation and enable Product in the mode selector.
Read MoreUse dsh-agent-teams to orchestrate DeepSeek Harness sessions into an intelligent agent team
`dsh-agent-teams` is a DeepSeek Harness community plugin maintained by NanmiCoder. It turns the current session into a team leader, allows spawning resumable sub-Agents by role, enables multi-agent collaboration with dependency-bound tasks, direct email access and a shared scheduler, and provides an activity panel on the web interface. This article verifies the installation command, ten tools, status directory and known limitations based on the directory page, GitHub README and usage.md. It is licensed under MIT, and you need to check the source code before installation; the plugin runs with the permissions of the current dsh process.
Read MoreY Combinator Open-Sources QM: Company-Grade Multi-Person Collaborative Agent Workspace
On July 31, 2026, Y Combinator open-sourced its internal multi-agent framework QM (Quartermaster) under the MIT license, and it quickly gained over 10,000 stars on GitHub. QM is a work-agent infrastructure for entire companies, supporting both Slack and Web terminals. Each employee and channel has isolated memory, sandboxes and permissions, and can switch Harness among Pi, OpenCode, Codex and Claude Code. This article sorts out its architecture, Strict/Auto/Dangerous security policies and qm init self-hosted deployment method, for teams evaluating enterprise-level agent operating systems for reference.
Read More