ruflo-swarm-sparc
Skill1 starsChanged 10 days agoMITView on GitHub
Integrates ruvnet/ruflo (Claude Flow), GSD, BMAD-Method, MetaGPT, ECC, and GitHub Spec Kit orchestration rules: Mission Proposal Gate, Resource Claims Board, SPARC workflow, Anti-Context-Rot, and 4-Step Project Memory.
In this repository45 other files it gives its agents.
- ask-mattSkill
- audit-context-buildingSkill
- codebase-designSkill
- code-eleganceSkill
- code-reviewSkill
- diagnosing-bugsSkill
- differential-reviewSkill
- dispatching-parallel-agentsSkill
- domain-modelingSkill
- elegant-delivery-docsSkill
- grill-meSkill
- grill-with-docsSkill
- handoffSkill
- impeccableSkill
- implementSkill
- implement-specSkill
- improve-codebase-architectureSkill
- karpathy-guidelinesSkill
- mcp-builderSkill
- ponytail-auditSkill
- ponytail-debtSkill
- ponytail-gainSkill
- ponytail-helpSkill
- ponytail-reviewSkill
- ponytailSkill
- postgres-best-practicesSkill
- prior-art-researchSkill
- prototypeSkill
- prSkill
- react-best-practicesSkill
- receiving-code-reviewSkill
- researchSkill
- retroSkill
- scrapling-officialSkill
- setup-matt-pocock-skillsSkill
- setup-pre-commitSkill
- subagent-prompt-architectureSkill
- tddSkill
- to-specSkill
- to-ticketsSkill
- triageSkill
- using-git-worktreesSkill
- verification-before-completionSkill
- wayfinderSkill
- webapp-testingSkill
The file
Files
/ruflo-swarm-sparc — 多源开源框架融合调度协议(Ruflo + GSD + BMAD + MetaGPT + ECC + Spec-Kit)
本技能浓缩了 GitHub 六大主流多 Agent 工程框架的核心机制,适用于任何编程语言与项目类型:
一、 智能分级与「先报方案卡再开工」门禁(Ruflo Mission Proposal + BMAD Right-Sized Process)
- Tier 0(单兵直通,免确认、免派子代理):
- 触发条件:纯查询问答、简单环境操作、或单文件 ≤20 行的明确微调(GSD #1979 内联执行原则)。
- 行为:主理官直接查证、执行并实测汇报,零多 Agent 调度开销。
- Tier 1~3(提案先行门禁:Proposals, Not Orders):
- 触发条件:凡涉及新功能开发、修改 ≥2 个文件、架构/存储/状态变更、或存在多种技术实现路径。
- 行为:主理官严禁盲目直接改代码!必须先完成「本地代码静默扫描 + 外部优秀开源项目检索(新功能/选型时)」,然后向用户输出一张结构清晰的 《任务理解与开工方案卡(Mission Proposal Card)》:
- 🎯 转化后的工程目标(将用户的大白话转化为明确的功能、边界与隐性防呆需求);
- 🔍 现成开源项目/原生方案参考(列出搜到的 1~3 个高星开源项目、成熟库或项目内既有组件);
- ⚖️ 关键设计取舍与推荐项(1~2 个核心决策点,附优劣对比与明确推荐项);
- 🤝 拟派专家、文件独占锁(Resource Claims)与验证计划。
- 放行条件:收到用户确认(如“可以”、“按推荐开始”)后,立即派发专家高效执行。
二、 SPARC 五阶段研发流水线与渐进式上下文交接(Ruflo SPARC + MetaGPT SOP + BMAD Handoff)
- S — Specification(需求解码与边界锁定):主理官解码用户大白话,补齐隐性体验需求(耗时操作进度提示、防并发重入互斥锁、向后兼容、一键式操作)。
- P — Prior-Art & Pseudocode(开源对标与最小原型设计):浮光检索 GitHub 同类优秀开源实现与官方文档,提炼可直接复用的组件与避坑清单。
- A — Architecture & Call-Graph Matrix(全调用链建模与规格拆票):跨模块/多文件任务由璇玑扫描全仓所有读、写、删、分页与定时任务调用点,输出《全调用点接线核对表》与曳光弹工单。
- R — Refinement(TDD 深模块与优雅代码实现):巧夺与玲珑在独立文件锁(Claims)保护下按
/code-elegance与/tdd实现,澄澈按>80% 置信度门槛做安全与精简审查。 - C — Completion(实证门禁与杂志级文档交付):跑通全部测试与构建门禁,输出结构化交付卡,并为新模块或架构改造自动生成独立
.md接线指南。
三、 资源独占锁(Ruflo Resource Claims Board)与防上下文腐烂(GSD Anti-Context-Rot)
- 文件写锁(One Owner Per Resource):并行派发多个子专家时,主理官必须在委派提示词中明确每个专家独占修改的文件列表;严禁两个活跃子专家同时写同一个文件。只读审查员(Reviewer / Verifier)严格禁止修改任何文件。
- 主理官极薄上下文(Thin Orchestrator):主理官不把成千上万行源码读进自己的主会话,而是将文件检索、全调用链扫描和重度编码委派给拥有独立 200k 干净上下文的子专家,彻底避免长会话上下文腐烂(Context Rot)。
四、 四步项目记忆蒸馏(Ruflo 4-Step Intelligence Pipeline)
- RETRIEVE(检索) → JUDGE(评估有效性) → DISTILL(提炼通用模式) → CONSOLIDATE(固化至当前项目
GLOSSARY.md/docs/adr//.scratch/),让当前项目越用越聪明,同时保持专家团自身提示词 100% 纯净通用。
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
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.

