anthropics / anthropics/claude-agent-sdk-python

Claude says that I send an empty message but I didn't

オープン
#710 コメント 0 件 リアクション 1 件 担当者 0 名 GitHub で見る
主要言語
Python
スター
8.1k
フォーク
1.3k
平均マージ
2日 31分
マージ済み PR(30日)
1

説明

I had this a few times already, I sent Claude on some long task (with sub-agents), and then at some point, it does not give me the result, but it tells me:

```It looks like your message came through empty! Did you mean to ask something? 😊```

Then, when I look at the logs, I noticed that the sub-agent did complete and return the information the main agent would need, but it didn't use it. Then there is also a `ThinkingBlock` that says:

```
The user didn't ask a question. They just triggered a system reminder. Let me wait for their actual question or message.
```

I have no idea how this happens.

I do see a `TaskNotificationMessage(subtype='task_notification' ... 'status': 'completed'` **before** the `ToolResultBlock` comes in, maybe that's the issue?

コントリビューションガイド

このリポジトリのコントリビューションガイドは索引されていません

評価

この issue はまだ評価されていません。

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。