Summary
Webchat is exposing internal AI prompt context to users. When sending messages via webchat, the following metadata block is visible to users:
Conversation info (untrusted metadata):
{
"message_id": "xxx",
"sender": "openclaw-control-ui"
}
### Steps to reproduce
1. Use webchat channel
2. Send a message
3. The "Conversation info (untrusted metadata):" block appears in the chat
### Expected behavior
This metadata is internal context for the AI (see `buildInboundUserContextPrefix` in `src/auto-reply/reply/inbound-meta.ts`) and should NOT be displayed to users.
### Actual behavior
- This started appearing recently
- The metadata is generated in `src/auto-reply/reply/inbound-meta.ts` as part of the AI prompt context
### OpenClaw version
2026.2.14
### Operating system
docker
### Install method
_No response_
### Logs, screenshots, and evidence
```shell
Impact and severity
No response
Additional information
No response
Summary
Webchat is exposing internal AI prompt context to users. When sending messages via webchat, the following metadata block is visible to users:
Conversation info (untrusted metadata):
{ "message_id": "xxx", "sender": "openclaw-control-ui" } ### Steps to reproduce 1. Use webchat channel 2. Send a message 3. The "Conversation info (untrusted metadata):" block appears in the chat ### Expected behavior This metadata is internal context for the AI (see `buildInboundUserContextPrefix` in `src/auto-reply/reply/inbound-meta.ts`) and should NOT be displayed to users. ### Actual behavior - This started appearing recently - The metadata is generated in `src/auto-reply/reply/inbound-meta.ts` as part of the AI prompt context ### OpenClaw version 2026.2.14 ### Operating system docker ### Install method _No response_ ### Logs, screenshots, and evidence ```shellImpact and severity
No response
Additional information
No response