agentleFS
Binbin-miaoBinbin-miao/workbuddy-expert-team/skills/dispatching-parallel-agents/SKILL.md

dispatching-parallel-agents

Skill1 starsChanged 10 days agoMITView on GitHub

Autonomous Wave-Based Parallel Multi-Agent Orchestration (synthesizing obra/superpowers dispatching-parallel-agents, GSD wave execution, LangGraph fan-out/fan-in supersteps, and Ruflo resource claims). Automatically spawns concurrent subagents across orthogonal domains without user prompting.

190 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use chinese-code-review, chinese-commit-conventions and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Leodorareluctant259Skill

    chinese-code-review

    中文代码审查规范——在保持专业严谨的同时,用符合国内团队文化的方式给出有效反馈

    Also in 8 projects that use this

  • Leodorareluctant259Skill

    chinese-commit-conventions

    中文 Git 提交规范 — 适配国内团队的 commit message 规范和 changelog 自动化

    Also in 8 projects that use this

  • obraSkill

    diagnosing-superpowers

    Use when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.

    Also in 19 projects that use this

  • kntismSkill

    condition-based-waiting

    Use when tests have race conditions, timing dependencies, or inconsistent pass/fail behavior - replaces arbitrary timeouts with condition polling to wait for actual state changes, eliminating flaky tests from timing guesses

    Also in 3 projects that use this

  • xfstudioSkill

    context-driven-development

    Use this skill when working with Conductor's context-driven

    Also in 3 projects that use this

  • EvezArtSkill

    analytics-dashboard

    User behavior tracking and analytics dashboards

    Also in 3 projects that use this

  • +3 more from Leodorareluctant259

The file

Files
SKILL.md

/dispatching-parallel-agents — 自主波次并发调度与多实例交叉验证协议

融合 obra/superpowers(dispatching-parallel-agents & subagent-driven-development)、gsd-build/get-shit-done(Wave-Based Parallel Execution)、LangGraph(Fan-Out / Fan-In Supersteps & Reducers)与 ruvnet/ruflo(Resource Claims Board):

一、 核心原则:无需用户催促,主理官按任务矩阵自主并发(Autonomous Fan-Out)

主理官(天枢)严禁被动等待用户输入“积极调用你的团队成员”或“用多个 Agent 交叉验证”!收到任务后,自动按以下决策矩阵决定单兵还是并发扇出(Fan-Out):

任务特征调度模式并发行为规范
Tier 0:微任务<br>纯问答、单条简单命令、单文件 ≤20 行明确微调单兵内联直通(0 子代理)主理官直接完成并实测交付,避免为小事拉起多 Agent 的通信延迟
Tier 1~3:审查仓库 / 排查复杂 Bug / 性能优化 / 新功能开发 / 跨模块改造强制三波次自主并发(Multi-Wave Fan-Out)主理官必须在同一回复回合内同时发出多个 Agent 工具调用,让多个子专家在各自干净的上下文窗口中真正并行工作

二、 三波次并发执行架构与同步屏障(3-Wave Execution & Barriers)

[Wave 1: 并行只读侦察波 (2~5 个并发 Agent 同回合发出)]
  ├─► 浮光 (WEB 对标检索):外部开源方案、上游仓库差异、官方文档/API 取证(只读,禁跑编译)
  ├─► 璇玑 (GRILL 领域建模):全仓读写/删除/分页调用链扫描、并发锁与崩溃原子性建模(只读,禁跑编译)
  ├─► 澄澈 (LEAN 极简代码):安全爆炸半径、静默吞错、性能热点与版本差分回归审计(只读,禁跑编译)
  ├─► 玲珑 (UI/UX 审美总监):[含界面时] 组件状态生命周期、异步监听器泄漏与可访问性审查(只读,禁跑编译)
  └─► 巧夺 (TDD 深模块研发):【唯一门禁实跑员】执行编译/类型检查/测试套件,猎捕实现级 Bug
        │
        ▼ ════════════ [Wave Barrier 1:全员回传收齐 → 主理官天枢汇总对账] ════════════
  [向用户呈交《任务理解与开工方案卡》──► 等待用户拍板确认]
        │
        ▼ ════════════ [Wave Barrier 2:用户确认放行] ════════════════════════════════
[Wave 2: 按模块独占写锁并行实现波 (1~3 个并发 Agent,按目录正交切分)]
  ├─► 巧夺-Core:独占修改后端核心模块目录(Resource Claim A)
  ├─► 玲珑-UI:独占修改前端组件与交互层目录(Resource Claim B)
  └─► 巧夺-Host:独占修改宿主适配/CLI/路由接线层(Resource Claim C)
        │
        ▼ ════════════ [Wave Barrier 3:实现全部收口] ════════════════════════════════
[Wave 3: 同角色多实例双盲交叉验证波 (2~3 个并发只读 Verifier 同回合发出)]
  ├─► 澄澈-Reviewer:静态差分审查 git diff(>80% 置信度过滤,禁跑重型编译)
  ├─► 巧夺-Verifier-A:【唯一门禁复跑员】从干净状态重跑全套编译、测试与静态检查门禁
  └─► 巧夺-Verifier-B:测试真实性与对抗/变异验证(强制在 /tmp 或独立 Worktree 中进行,严禁污染主工作树!)

三、 并发调度的五条防踩踏铁律(Five Anti-Chaos Rules)

  1. 同一回合扇出(Single-Turn Fan-Out):凡属同一波次且互不依赖的子任务,必须在同一个模型回复回合中同时发出多个 Agent 调用,绝不一个接一个串行等待。
  2. 编译锁单点化(Single Gate-Runner Discipline):在 Rust(cargo)、C++、Go、Java 或大型前端工程中,多个子 Agent 同时运行编译/构建会争抢编译目录锁导致长达数十分钟的挂起!因此同一波次中最多只允许指派 1 个 巧夺 实例运行编译/构建/全量测试,其余并发专家必须在委派 Prompt 中被明确要求:“全程静态阅读(grep/read/diff),禁止运行重型编译或测试命令”。
  3. 波次屏障(Strict Wave Barrier):在 Wave 1 的只读调研 Agent 全部返回之前,绝对禁止提前启动 Wave 2 的代码写入(防止实现写完才发现调研找出了架构级漏项而返工)。
  4. 主理官禁手与文件独占锁(Orchestrator Hands-Off & Disjoint Claims):
    • Wave 2 并行实现时,不同子 Agent 的可写文件集合必须两两互不相交(Disjoint File Sets),并在启动前锁定模块间接口契约;
    • 当任何实现或审查子 Agent 正在后台运行时,主理官(天枢)绝对禁止在主工作树上私自修改业务代码,防止覆盖子 Agent 的在途工作或导致审查行号失效!
  5. 沙箱化对抗与变异测试(Sandboxed Mutation Testing):
    • 当派出多个 巧夺 或 澄澈 实例做同角色多实例盲测或变异测试(验证“把生产代码改坏时测试能否变红”)时,严禁在主工作树上直接改写文件或执行 git checkout!一切破坏性探针与变异实验必须在系统临时目录(/tmp/...)或独立 Git Worktree(.worktrees/...)中执行。

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.