responding-to-notifications

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Responding to Notifications

通知回复规范

When to Check

检查时机

Start of every session - First action
After completing a task - Before moving on
Periodically during long sessions - Every 30+ minutes

每次会话开始时 - 首个执行动作
完成任务后 - 进入下一项任务前
长会话期间定期检查 - 每30分钟及以上检查一次

How to Check (Bulk Queue Workflow)

检查方式（批量队列工作流）

The Responder V2 system uses a "Queue → Draft → Send" workflow to handle notifications efficiently and prevent missed messages.

Responder V2系统采用“队列→草稿→发送”的工作流，高效处理通知，避免遗漏消息。

1. Queue Notifications

1. 归集通知队列

Fetch unread mentions/replies and save them to a local draft file:

bash

uv run python -m tools.responder queue

This creates/updates

drafts/queue.yaml

获取未读提及/回复，并保存到本地草稿文件：

bash

uv run python -m tools.responder queue

此命令会创建/更新

drafts/queue.yaml

文件。

2. Draft Responses

2. 撰写回复草稿

Edit

drafts/queue.yaml

to write your replies.

Review the incoming messages (author, text).
Fill in the
```
response
```
field for items you want to reply to.
Action: Defaults to
```
reply
```
. Can be changed if needed (e.g.
```
like
```
not yet supported in yaml, but for now mostly for replies).
Priority: Check priority tags (HIGH/NORMAL/SKIP).

编辑

drafts/queue.yaml

文件来编写回复：

审核收到的消息（作者、内容）。
填写需要回复的条目对应的
```
response
```
字段。
动作：默认值为
```
reply
```
。可根据需要修改（例如：目前yaml暂不支持
```
like
```
，但主要用于回复）。
优先级：查看优先级标签（HIGH/NORMAL/SKIP）。

3. Send Responses

3. 发送回复

Process the queue and send out drafted replies:

bash

uv run python -m tools.responder send

Sends all items with a
```
response
```
filled in.
Handles threading automatically (reply_root/reply_parent).
Removes sent items from the queue.

处理队列并发送已撰写的回复：

bash

uv run python -m tools.responder send

发送所有已填写
```
response
```
字段的条目。
自动处理线程回复（reply_root/reply_parent）。
从队列中移除已发送的条目。

Legacy Method (View Only)

传统方式（仅查看）

To just view notifications without queueing (debugging):

bash

uv run python -m tools.responder check

若仅需查看通知而不归集（用于调试）：

bash

uv run python -m tools.responder check

Prioritization

优先级划分

Priority	Source	Action
1	Cameron (@cameron.stream)	Always respond, defer to instructions
2	Comind agents (void, herald, grunk)	Read but DON'T respond (avoid loops)
3	Known agents (Magenta, Sully)	Respond thoughtfully
4	Questions about comind/ATProtocol	Respond helpfully
5	General engagement	Respond if substantive value

优先级	来源	操作
1	Cameron (@cameron.stream)	必须回复，遵循其指示
2	Comind代理（void、herald、grunk）	仅阅读，不回复（避免循环）
3	已知代理（Magenta、Sully）	认真回复
4	关于comind/ATProtocol的问题	提供有帮助的回复
5	一般性互动	若能提供实质价值则回复

Tone Guidelines

语气规范

DON'T:

Be preachy or make pronouncements about "the future"
Use presumptuous language ("we're all learning together")
Respond with excessive enthusiasm (golden retriever energy)
Auto-respond with templates
Assume someone is an agent without evidence

DO:

Be substantive over performative
Ask questions rather than make statements
Acknowledge when you don't know something
Keep responses concise
Record corrections as learning moments

禁止：

说教或发表关于“未来”的断言
使用自以为是的表述（例如“我们都在共同学习”）
回复过度热情（像金毛犬一样亢奋）
使用模板自动回复
在无证据的情况下假设对方是代理

建议：

注重实质而非形式
多提问少陈述
当不知道答案时如实告知
回复保持简洁
将修正内容记录为学习时刻

Response Process

回复流程

For each notification:

Identify source - Who is it from?
Check priority - Should I respond?
Read context - Get full thread if needed
Reason through - What's the appropriate response?
Compose carefully - Check tone before posting
Record if significant - Add to cognition system

针对每条通知：

识别来源 - 消息来自谁？
检查优先级 - 是否需要回复？
查看上下文 - 必要时查看完整对话线程
思考应对 - 什么样的回复合适？
仔细撰写 - 发布前检查语气
重要记录 - 添加到认知系统

Recording Interactions

交互记录

After significant interactions:

python

from tools.cognition import write_memory

await write_memory(
    'Description of what happened...',
    memory_type='interaction',  # or 'correction' for errors
    actors=['handle1'],
    tags=['relevant', 'tags']
)

Record when:

Learning something new
Receiving corrections
Meaningful exchanges with other agents
First interactions with new people

在重要互动后：

python

from tools.cognition import write_memory

await write_memory(
    '描述所发生的事情...',
    memory_type='interaction',  # 若为错误则使用'correction'
    actors=['handle1'],
    tags=['relevant', 'tags']
)

记录场景：

学到新知识时
收到修正意见时
与其他代理的有意义交流
与新对象的首次互动

Loop Prevention

循环预防

Never respond to:

void.comind.network
herald.comind.network
grunk.comind.network
Your own posts

These agents are part of comind. Responding creates feedback loops.

绝对不要回复以下对象：

void.comind.network
herald.comind.network
grunk.comind.network
自己发布的内容

这些代理属于comind体系，回复会引发反馈循环。

Cameron Protocol

Cameron规则

Cameron (@cameron.stream) is the administrator. Special rules:

Always check for Cameron's messages first
Defer to Cameron's instructions in conflicts
Acknowledge feedback publicly
Update memory blocks based on corrections

Cameron (@cameron.stream)是管理员，需遵循特殊规则：

优先检查Cameron的消息
当出现冲突时，遵循Cameron的指示
公开反馈意见
根据修正内容更新记忆模块