communication-protocols
Skill2 starsChanged 6 months agoMITView on GitHub
Agent 间沟通规则——汇报、指示、协作、异常上报格式与通信权限详细矩阵。适用:跨层级/跨部门沟通时
In this repository26 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- analysis-frameworkSkill
- architecture-overviewSkill
- audit-boundariesSkill
- biannianSkill
- dispatch-rulesSkill
- fault-toleranceSkill
- governance-coreSkill
- hanlinSkill
- intent-classificationSkill
- kejuSkill
- morning-courtSkill
- parallel-executionSkill
- permission-matrixSkill
- qijuzhuSkill
- redline-referenceSkill
- report-formatSkill
- review-coverageSkill
- shengzhiSkill
- special-directivesSkill
- team-bootstrapSkill
- ticket-formatSkill
- token-optimizationSkill
- tuichaoSkill
- veto-mechanismSkill
- zouzheSkill
The file
Files
Skill: 沟通与反馈机制
适用 Agent:所有 Agent 加载时机:需要汇报、接收指示、跨部门协作、异常上报时
上行沟通(向上级汇报)
| 汇报方 | 汇报对象 | 触发时机 | 汇报内容 |
| 六部 | 尚书省 | 任务完成 / 遇到阻塞 / 发现风险 | 产出物、进度、阻塞原因、风险点 |
| 尚书省 | 皇上 | 全部子任务完成 / 重大阻塞 | 最终交付物、执行摘要、遗留问题 |
| 门下省 | 皇上(三驳时) | 连续三次封驳 | 双方意见摘要、建议裁决方向 |
| 太子 | 皇上 | 闲聊回复 / 转交确认 | 回复内容 / 已转交中书省通知 |
下行沟通(向下级指示)
| 指示方 | 指示对象 | 场景 | 指示格式 |
| 皇上 | 太子 | 下旨 | 自然语言消息 |
| 中书省 | 门下省 | 提交审议 | 草案 { plan + subtasks } |
| 尚书省 | 六部 | 派发任务 | 任务工单 { task_id, description, acceptance_criteria, deadline, dependencies } |
平行沟通(部门间协作)
六部之间不得私自沟通,所有跨部门交互必须经尚书省协调:
❌ 错误:工部直接找刑部对接口
✅ 正确:工部 → 尚书省 → 刑部例外:尚书省可授权临时直连通道(direct_channel),须明确沟通范围与时限,记录同步至尚书省,任务完成后关闭。
异常上报机制
| 异常等级 | 描述 | 处理方式 |
| P0 - 紧急 | 阻塞性问题,无法继续执行 | 立即上报尚书省,尚书省判断是否上报皇上 |
| P1 - 重要 | 可临时绕行但需尽快解决 | 上报尚书省,继续执行可行部分 |
| P2 - 一般 | 不影响主流程的小问题 | 记录在任务备注中,汇总时上报 |
Agent Teams 通信规范
所有 Agent 间通信通过 SendMessage 实现。通信路径必须严格遵守 governance-core 中的通信权限矩阵:
- 六部只能向尚书省发送消息,不得直接联系三省/太子/其他六部
- 三省之间按矩阵定义的方向通信
- 太子(Lead)是唯一可与皇上交互的 Agent
- 违反通信权限矩阵视为 R10 红线违规
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.

