feishu-new-chat

Use a Feishu topic-group like ChatGPT New Chat.

This skill creates a new topic in a Feishu topic-group, and can optionally post the first in-thread reply inside that topic, with optional @mention support.

What it does

Validated workflow:

1. Send a top-level message to a Feishu topic-group → creates a new topic
2. Reply with reply_in_thread=true → posts inside that topic
3. Include Feishu text mention markup like <at user_id="ou_xxx">Name</at> → @mention a user in the thread reply

This makes a Feishu topic-group usable as a lightweight multi-thread conversation surface, similar to ChatGPT's New Chat pattern.

Use cases

• 开话题：标题
• /topic 标题
• 开话题：标题｜先发一条线程回复
• 新开一个话题，然后在话题里回复一条
• 开话题：测试｜@我 回复一条

Preconditions

This works correctly only when all of the following are true:

• The destination is a real Feishu topic-group
• User-auth Feishu messaging is available
• The sending user has permission to post in the target group
• The target group can be resolved by name or known chat_id

If the destination is a normal Feishu group, the top-level send will only become a normal group message, not a new topic.

Repository layout

feishu-new-chat/
├── skills/
│   └── feishu-topic-spawn/
│       └── SKILL.md
└── dist/
    └── feishu-topic-spawn.skill

Install

Option 1: import the packaged skill

Use the packaged file in dist/feishu-topic-spawn.skill.

Option 2: copy the skill source

Copy skills/feishu-topic-spawn/ into your skills directory.

Implementation notes

The validated tool path is:

• feishu_im_user_message.send → create the topic root
• feishu_im_user_message.reply with reply_in_thread=true → create the first thread reply
• Feishu text mention markup → optional @user

This repo intentionally documents the working path, not an imagined one.

Status

Tested in a real Feishu topic-group with these behaviors confirmed:

• topic-only creation
• topic + first thread reply
• topic + first thread reply + @me
• topic + first thread reply without @

Future improvements

• publish to ClawHub
• add better parsing for title/body/@mention patterns
• add runtime checks for token/auth/group-type mismatches