Cavan-Ou/hermes-dsh-collab 预览 preview

Cavan-Ou/hermes-dsh-collab

经过实战检验的多Agent协作手册,适用于DeepSeek Harness:模型层路由、规格纪律、git单写者规则——作为可安装技能。在多Agent管线中运行DSH的实战协作规范。

Project Overview项目介绍

This is a native skill plugin built explicitly for DeepSeek Harness (DSH), designed to integrate DSH as a task executor into Hermes AI orchestration pipelines. It enables automated task dispatch, execution, and verification, with quality gates that do not rely on AI self-reporting to catch hidden errors. It supports two installation methods: you can install it directly via the official DSH plugin command, or clone the repository manually and copy the skill folder to your local DSH skills directory. After installation completes, the skill loads automatically on your next DSH session, and you can run three quick checks to verify it works correctly.

This plugin targets developers who need to run long, unattended multi-agent AI pipelines for software projects. It follows a clear separation of responsibilities that has been tested in a real 14-day production pipeline with 30 total commits. The Hermes orchestrator writes the task dispatch spec, DSH executes the task following pre-defined rules, and Hermes handles post-execution quality verification. You only need to step in when a quality gate fails, which eliminates the need for constant manual oversight of routine task stages.

This plugin requires DSH version 0.1.x running on a headless profile, and it has been fully verified to work on DSH 0.1.0-rc.6. It is released under the open-source MIT license, so you can use, modify, and distribute it freely per the terms of the license. There is one known limit that the plugin explicitly documents: the upstream DeepSeek Pro model channel may be unavailable at times. It also lists 12 common pitfalls encountered during real-world use, complete with symptoms, root causes, and step-by-step fixes for each issue.

这是一个专为DeepSeek Harness(DSH)打造的原生技能插件,用于将DSH接入Hermes多AI代理编排流水线,实现任务的自动分发、执行与验证。它支持两种安装方式:通过DSH官方插件命令dsh plugin add安装,也可手动克隆仓库后复制到DSH技能目录完成安装,安装后重启DSH会话即可自动加载。它定位于填补编排器与执行器之间的操作规范缺口,减少因规则缺失导致的返工。

该插件适用于需要无人值守运行长期多AI任务流水线的开发者,完整工作流程为Hermes编写任务分发规范,DSH按预定义规则执行任务,Hermes负责后续质量验证,开发者仅需在质量门检测失败时介入处理。插件内置了经过14天实际项目验证的规则,包括模型路由分层、规范三原则、Git单写者规则等多项约束,避免不规范操作。

该插件依赖0.1.x版本的DSH,需使用无头配置文件运行,当前验证兼容的DSH版本为0.1.0-rc.6,采用MIT许可协议免费开源。需要注意的是,当前DeepSeek Pro模型通道上游可能不稳定,插件已明确标注该限制,还记录了实际使用中遇到的12种常见陷阱及对应解决方案。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 7 stars - very few users, little community feedback星标只有 7,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile headless add github:Cavan-Ou/hermes-dsh-collab

把 Cavan-Ou/hermes-dsh-collab 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

hermes-dsh-collab

Hook DeepSeek Harness into your Hermes pipeline: automated dispatch, execution, and verification — with quality gates that don't trust self-reports.

MIT DSH Awesome DSH Plugin dsh-plugin Hermes

简体中文版见 README.zh.md

Your AI assistant can work all day. The question is whether you can walk away.

This skill makes it safe to: Hermes writes the dispatch spec, DSH executes it, Hermes verifies it — and you only step in when a quality gate actually fails. Distilled from a real 14-day pipeline (30 commits, 7/7 stages shipped without rework).

Dependencies: DSH 0.1.x (headless profile) + any orchestrator agent (built and tested with Hermes).


Why this exists

Hermes is your personal assistant. DSH is a capable executor. The gap is the operating contract between them: what a good dispatch spec looks like, which model tier to use for which stage, who is allowed to commit, and how verification actually happens.

Most pipelines skip that contract and pay for it in rework. Recent work (COPE) shows planner/executor separation works — but only when the executor's behavior is controlled. This skill encodes those controls, from a pipeline where they were tested:

  • Model-tier routing — Flash (reasoning: max) for routine stages and for looking at images (the main model is natively multimodal — no vision tier, no vision patch), Pro for multi-file refactors and long synthesis (its channel is currently down upstream; the table says so instead of pretending otherwise). When in doubt: try Flash once — rework means escalate.
  • Spec 3 iron rules — Plan first · test first (TDD red→green) · scope declaration. A spec missing any of these is not shippable.
  • Git single-writer — only the orchestrator commits. The executor never touches git, so history stays linear and auditable.
  • Quality gates owned by the orchestrator — full test suite + build + diff-vs-scope audit + real browser walkthrough. Self-reports are not evidence.
  • Write-back via cwd — cd <project> && dsh --profile headless "task" writes straight back. No /tmp mirrors, no patch handoffs.
  • Pitfalls with receipts — every entry is a real incident: patch config replaces whole sections (not deep-merge), read chokes on binaries while read_image needs the model to declare image input, stale backend processes invalidate green tests…

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev chicheng-cron 下一个 Next dsh-tidewatch →