Apache-2.0 许可、面向 DeepSeek Harness 与 Codex 的多智能体编排插件，把一个目标拆成隔离的并行任务通道，让一次对话无需承载整个项目。
DSH 适配
原生运行时
作者声明
安全审计
未审计
最后核验
2026-09-01
许可证
Apache-2.0
01它能帮你完成什么？
Orchestrate a large multi-part engineering goal (e.g. refactor authentication end to end) as independent parallel tasks
Independent top-level tasks run in parallel with separate contexts, waiting only on real dependencies, with results merged back together
Developers using DeepSeek Harness (dsh) or Codex who want to split one big goal into isolated, parallel task lanes
Let each task drive its own subagents, tools, Skills, or MCPs while keeping contexts isolated
Each task executes with its own scoped subagents, tools, Skills, and MCPs, recursively, without contaminating other tasks
AI coding agents and developers needing recursive multi-agent workflows without a single growing conversation
02如何接入 DeepSeek Harness？
作者未说明
03DSH 适配与能力边界
First-class All in Flash DSH plugin: a DeepSeek Harness (dsh) / Codex multi-agent orchestration layer that breaks one goal into independent, parallel task lanes.
Parallel task orchestration
one large goal or prompt→independent top-level tasks executed in parallel with isolated contexts, dependencies respected, results merged
Per-task subagents, tools, Skills, and MCPs
each top-level task→task-scoped subagents, tools, Skills, and MCPs executed recursively
each task maintains its own separate context, isolated from other tasks
04适合谁？何时不该用？
适合
- Developers using DeepSeek Harness (dsh) or Codex who want to split one big goal into isolated, parallel task lanes
- AI coding agents and developers needing recursive multi-agent workflows without a single growing conversation
不适合
- Without task isolation, unrelated work contaminates other work and earlier constraints become easier to forget; All in Luna keeps each task's context separate to avoid this.
05兼容性、维护与安全提示
- Without task isolation, unrelated work contaminates other work and earlier constraints become easier to forget; All in Luna keeps each task's context separate to avoid this.
Apache-2.0 · latest release allinflash-v0.2.0 (2026-08-14), last push 2026-09-01
06常见问题
All in Luna 解决了什么问题？
把整个项目放在一次 AI 对话里会让上下文不断增长，无关工作互相污染，并可能因一个局部阻塞而拖垮整个流程。All in Luna 让每个任务保持独立上下文，并并行运行相互独立的任务。
它与 DeepSeek Harness 是什么关系？
它以 All in Flash DSH 插件形式发布，专为 DeepSeek Harness（dsh）和 Codex 打造，在普通子智能体之上增加一层多智能体编排。
每个任务还能使用子智能体和工具吗？
可以。每个任务都可以使用自己的子智能体、工具、Skills 或 MCP——任务之间并行，任务内部可递归。
它是开源的吗？
是的，仓库采用 Apache-2.0 许可。
07相关的 DSH 工作流
archify
作者 tt-a1i
为编码智能体生成美观可验证的架构图、时序图与数据流图，输出自包含 HTML，支持动效与清晰导出。
openviking
作者 volcengine
为 AI 智能体打造的自进化上下文数据库，统一智能体记忆、知识 RAG 与技能。
nocobase
作者 nocobase
开源的 AI + 无代码应用搭建平台：AI 在久经生产验证的底层基础设施与所见即所得的无代码界面之上工作，帮你快速构建 CRM、ERP 等业务系统，兼顾速度与可靠性。
learn-harness-engineering
作者 walkinglabs
Harness 工程新手教程，从 0 到 1 系统学习智能体工作流框架。
08数据与来源
It turns the work into independent top-level tasks
页面基于项目公开文档、仓库元数据和 DSH Plugins 的结构化解析生成；最后核验于 2026-09-01。发现错误？提交更正。
最佳 DeepSeek Harness 插件
从全目录挑出的 12 个值得优先安装的插件，覆盖各个分类。