AI管大师·五大护法：用 7 人 AI 战队承接中小企业管理协作
AI管大师·五大护法是SkillHub社区版ai-agent Skill（v2.1.13，@user_e23813cc/ai-guandashi-enterprise-v2），基于QClaw/OpenClaw，打包大内总管、管家婆与五位业务Agent共七个角色、YAML模板及七条跨护法协作工作流，强调人走不递补。适用于贸易电商、服务业、互联网科技与中小企业总部职能，支持企业微信与个人微信双渠道AI管理后台部署。文章说明核心能力、skillhub install安装步骤及Cursor等客户端skills目录，并给出典型启用顺序与使用注意。
Read MoreHermes 自进化协作系统：用原生工具做多 Agent 协作与技能沉淀
Hermes 自进化协作系统（@user_b2d76d7e/hermes-evolution-plus）是 SkillHub 上的 ai-agent 类 Skill，在 Hermes 原生工具 delegate_task、memory、skill_manage、todo 上融合 P0-P6 行为纪律、PM 协议、任务状态机、学习闭环、技能沉淀与自纠错及上下文瘦身，并附带 Python/Node 的 4 级压缩与信号检测脚本。本文基于官方页面说明介绍能力边界、SkillHub 安装命令与典型工作流，并说明发布方对比原生 Hermes 的 token/速度/成本自述及适用注意。
Read More可行性研究报告撰写 · 多Agent长报告：超长可研的分阶段协作 Skill
SkillHub 社区 Skill「可行性研究报告撰写 · 多Agent长报告」（@user_0c7f1934/long-report-agent，v1.0.6）面向 10+ 章节可研、项目建议书与实施方案。流程为 Phase 0 需求确认、Phase 1 规划大纲、Phase 2 最多 5 个子 Agent 并行撰写、Phase 2.5 跨章一致性审查、Phase 3 整合输出 docx（含多种封面、表格式目录与美化表格）。无需 API Key。安装可参考 skillhub.cn 文档或用 skillhub install 指定 Cursor/Claude 等 skills 目录。触发语包括「写可研」「可行性研究报告」「把章节汇总成 docx」等。
Read More玄机子：在 WorkBuddy 里用多 Agent 调度传统命理咨询
玄机子是 SkillHub 社区 Skill（@user_70b62916/573215058，v1.0.0，professional），无需 API Key。加载后 WorkBuddy 以玄机子角色，经 TeamCreate、Agent spawn、SendMessage 调度十三位命理宗师，覆盖八字、紫微、六爻、风水、择日等门类，用白话输出综合解读。文章说明定位、触发条件、官方安装提示与适用宿主限制，并给出可复现的提问方式。
Read More帝国创世纪：把 OpenClaw 一键升级为龙虾帝国标准架构
帝国创世纪（@user_f9be8d90/empire-genesis，v1.1.0）是 SkillHub 社区 ai-agent 类 Skill，无需 API Key。它将 OpenClaw 实例自动升级为龙虾帝国标准架构：建跨 OS StateDB、注入 Signal Protocol v2、补齐 research/execution/verification Agent、部署带心跳补丁的 Watchdog，并经冷链与铁猪验证后无痕清理。文章说明五项核心动作、SkillHub CLI 安装命令（含 Cursor/Codex 等 skills 目录）、以及「初始化帝国」等触发词用法与适用注意。
Read Morefeishu-agent-send：多 Agent 场景下的飞书互发消息
feishu-agent-send（@user_d1914364/feishu-agent-send，SkillHub v3.0.2，ai-agent）是社区发布的飞书多 Agent 通信 Skill。通过 ~/.feishu_agent_send/config.json 与 feishu_set_self、feishu_add、feishu_send 等 Python 工具，登记各 Agent 的 chat_id，并用 --deliver 生成 feishu_im_user_message 发送指引。文章说明 SkillHub 安装命令、Cursor/Claude/Codex skills 目录、典型配置流程及私聊群聊切换与常见错误处理，适合 OpenClaw 多 Agent 需在飞书互通知的场景。
Read More多Agent协作的小说创作系统：从灵感到成品的 Skill 流水线
SkillHub 社区技能「多Agent协作的小说创作系统」（@user_9aa92135/novel-writer-team，v1.0.1）面向小说与网文创作，用 8 个专业角色覆盖从灵感到成品的流程，并适配起点中文网、番茄小说、晋江文学城、知乎盐选、豆瓣阅读五种风格。默认自动推进，用户说「一步步来」可切换分步确认；无需 API Key。安装时向 AI 发送官方提示，按 skillhub.md 安装该坐标即可。
Read Moreacademic-paper-team：多 Agent 协作的学术论文写作 Skill
academic-paper-team 是 SkillHub 社区 ai-agent 类 Skill（v1.0.0，@user_cbe7a01d/academic-paper-team），在写论文、选题、文献综述等触发词下启动 Controller+四位专家+台账与五道闸门的有状态论文流程，融合 Sprint 契约与分离评估。本文基于官方页面说明架构、安装（skillhub 或安装文档提示词）及典型用法与适用注意。
Read More混合动态多 Agent 协作系统：按任务复杂度动态组队
SkillHub 社区 Skill「混合动态多 Agent 协作系统」（v1.2.1，@user_f076293a/dynamic-multi-agent-system）面向 OpenClaw 等支持 SKILL.md 的 Agent 环境。它自动识别单一、标准、创新三类任务，按需创建临时子 Agent 并行或串行协作，并提供三层质量检查、异常处理与可选流程固化。文章说明核心能力、skillhub 安装命令与各客户端 skills 目录，并给出官方文档中的任务描述示例与适用边界。
Read More中国专利申请文件多Agent并行审查器：并行审校并输出带修订的 docx
SkillHub 社区 Skill「中国专利申请文件多Agent并行审查器」（@user_dcf8e29c/checking-patdocx-cn-multi-agents，v1.1.3）面向中国专利申请 .docx/.doc，采用多 Agent 并行架构做检查、审阅与校对，并生成含修订追踪与批注的 docx 副本。适用于明确要求多 Agent 或并行审查的场景；不用于仅浏览文档、非 Word 格式或翻译需求。安装可参考 skillhub.cn/install/skillhub.md，用 skillhub install 并指定 Cursor 等客户端的 skills 目录。标签 requires_api_key=false。
Read More投标标书多 Agent 协作生成 Skill：六阶段流水线与 IMA 知识库优先
SkillHub 上的投标标书多 Agent 协作生成 Skill（v2.1）面向投标文件、技术/商务方案与响应矩阵写作。流程为总调度解析、IMA 知识库检索、并行章节编写（知识库优先并标记 AI 补全）、人工确认、合稿排版、五道硬门槛质检与五维评审，覆盖政策合规、技术匹配、商务匹配与格式检查。本文说明定位、核心能力与基于 skillhub.md 的安装方式，并给出典型协作步骤与使用注意。
Read More记忆管家：多 Agent 上下文与记忆文件的监控与复盘
记忆管家是 SkillHub 社区 ai-agent 类 Skill（v1.0.3，无需 API Key），面向多 Agent 长期运行：监控会话上下文与记忆文件磁盘占用并在超阈值时提示压缩，配合 crontab 每日增量扫描记忆、提炼偏好与待办并更新 MEMORY.md，安装后引导初始化目录与定时任务。本文基于官方页面与 SkillHub 安装文档说明安装坐标、CLI 命令及 Cursor/Claude Code/Codex 的 skills 路径，并归纳适用场景与限制。
Read Moredsh-round-table: Multi-Agent Collaborative Roundtable
`dsh-round-table` is a plugin for DeepSeek Harness designed to address context fragmentation and collaboration tracking issues in multi-agent development. It provides a centralized “round table” environment with features such as task card generation, visual workflows, shared resources, and independent secretary notes. To use it, install and enable the plugin in the DSH 0.2.0-rc.2 desktop version, then create a meeting, invite members, generate and publish task cards, and continuously follow up on collaboration. The plugin supports sequential or parallel orchestration while preserving each member’s original session context. Note that it is only available in the desktop version, and specific limitations apply to file handling and permission management.
Read MoreDeepSeek Harness Visual Multi-Agent Orchestration Plugin
The dsh-visual-workflow plugin provides a visual multi-agent workflow orchestration solution for the DeepSeek Harness (DSH) Web GUI. It supports drag-and-drop node editing, workflow and context connections, checkpoint-based resumption, scheduled triggering, and one-click publishing as an API service. The plugin features autonomous agent meta-orchestration capabilities, supports human-in-the-loop intervention, and has zero official package dependencies. It introduces the concept of "meta-orchestration," combining workflows, agent teams, and organizational memory; supports SQLite, MySQL, and PostgreSQL databases; and aims to address dynamic workflow orchestration challenges in complex tasks.
Read MoreDeepSeek Harness Plugin: An RTS Starship-Style Multi-Agent Orchestration Dashboard
dsh-plugin-stardeck is a plugin built for DeepSeek Harness, designed to address the chaos of multi-session management. It transforms the traditional “chatting with an agent” workflow into an RTS-style space-fleet command experience, where you issue orders to a fleet. The plugin’s core features include a 3D stellar view, four-stage lifeline management, front-line concept aggregation, and three levels of autonomy (Do Directly, First Mate Triage, and Review Plan First). It supports persistent ledger records and machine-verifiable checks, making it suitable for use cases such as weekly summaries and code review. The project was split/forked in 2026, with its core ported to a standalone daemon + MCP.
Read Moredsh-swarm: Autonomous Team Collaboration Plugin v0.4.0
dsh-swarm v0.4.0 is a multi-agent collaboration plugin for DeepSeek Harness, designed to address issues such as task loss, token waste, and latency. Its core features include resident agent processes, independent API configuration, artifact reference passing, a task ledger with watchdog-based automatic retries, and built-in online capabilities. Through a visual UI and toolchain, it supports multi-role task division and real-time monitoring for complex workflows. Compared with agent-teams, its independent processes and intelligent retry mechanism make it better suited for long-running and high-reliability scenarios.
Read Moredsh-ultramath: UltraMath Mathematical Modeling Contest Multi-Agent Solver Plugin
dsh-ultramath is a multi-agent solving plugin for DeepSeek Harness designed for mathematical modeling competitions, developed by Andiii208 and licensed under the MIT License. Through native Agent presets and skill packs, this plugin standardizes the mathematical modeling solution workflow into seven stages—"problem reading, framework, derivation, coding, verification, paper writing, and review"—thereby solving the fragmentation issue of single-turn prompt processes. Core features include providing the UltraMath master controller and four single-stage role presets (e.g., Mathematician, Engineer); integrating a built-in library of 33 models and paper templates to enable self-contained installation and idempotent local data synchronization. It establishes strict quality gates covering seven categories of controls: independent verification, review blockers, blind review mechanisms, and academic integrity, where any violation results in an immediate veto. Furthermore, it introduces a "digital freeze" mechanism to ensure the consistency between paper data and code.
Read MoreOne Command to Launch Agent Teams: Getting Started with DeepSeek Harness Plugin dsh-agent-teams
dsh-agent-teams is a DeepSeek Harness workflow plugin maintained by NanmiCoder that transforms the current session into a leader, creating resumable sub-agent teams. It splits tasks based on dependencies and automatically schedules them, complemented by a web activity panel. This article verifies the installation commands, core capabilities, and usage boundaries based on the SkillHub directory page and the official GitHub README, making it suitable for DSH users who need multi-agent collaboration without building a custom workflow engine.
Read MoreUsing dsh_workflow to make DSH's one-time multi-Agent scheduling into a recoverable workflow
`dsh_workflow` is a DeepSeek Harness workflow plugin maintained by icetomoyo (MIT license, package name: `@dsh-external/workflow`). Instead of replacing DSH's front-end one-off workflow tool, it supplements a process layer that supports generation, saving, governance, observation and recovery: including named capsules, persistent run storage, pause and resume execution, and QuickJS isolated execution. This article verifies the installation commands, slash usage and permission boundaries in accordance with the community directory page and GitHub repository.
Read MoreCursor Multi-Agent Swarm Rebuilds SQLite for $1339: Planner/Worker Hierarchy Is the Economics of Agents
In July 2026, Cursor released the Agent Swarm research: relying solely on 835 pages of SQLite documentation, without source code or network access, the multi-agent swarm rebuilt SQLite in Rust and passed all sqllogictest tests. The total cost of the hybrid solution planned by Opus 4.8 plus executed by Composer 2.5 is approximately $1,339, while the cost of the single-model GPT-5.5 solution is about $10,565. This article sorts out the Planner/Worker layered architecture, the new harness coordination mechanism, experimental data and the open-source minisqlite product, and analyzes the cost advantage of the multi-agent approach of "using strong models for planning and weak models for execution" as well as its enlightenment to engineering practice.
Read MoreCursor 3 Multi-Agent Architecture: Planner/Worker Division Reduces Costs by Up to 15 Times
In July 2026, Cursor unveiled its upgraded Agent Swarm: the cutting-edge Planner model splits tasks and formulates plans, while the low-cost Composer 2.5 Worker is responsible for execution. In a closed benchmark where it replicated a database from scratch using only the SQLite manual, all configurations of the new architecture achieved a 100% pass rate in sqllogictest. The total cost of the Opus+Composer combination was approximately $1,339, nearly 15 times lower than the roughly $10,565 cost of a single GPT-5.5 model; the Opus configuration’s code volume was reduced by about 85% compared to the old Swarm. This article outlines the tree-shaped division of labor, self-developed VCS, failure mode governance and model economics, and notes that while this result represents a Cursor self-developed experiment, caution is still needed for its deployment in production.
Read More