Bug

The web UI (Control UI chat interface) renders the raw security wrapper markup that OpenClaw injects around untrusted content, rather than stripping/hiding it.

Steps to Reproduce

1. Open the Control UI web dashboard
2. Navigate to a Discord channel that has a channel topic set
3. Start a chat session
4. Observe the raw markup visible in the chat interface

Expected Behavior

The <<<EXTERNAL_UNTRUSTED_CONTENT id="...">>> / <<<END_EXTERNAL_UNTRUSTED_CONTENT>>> tags should be stripped before rendering in the UI. They are an internal security mechanism and should never be user-visible.

Actual Behavior

The full markup block is rendered as bold text in the chat, e.g.:

<<<EXTERNAL_UNTRUSTED_CONTENT id="1909cdd23df75dd2">>>
Source: Channel metadata

UNTRUSTED channel metadata (discord)
Discord channel topic: ...
<<<END_EXTERNAL_UNTRUSTED_CONTENT id="1909cdd23df75dd2">>>

Additional Context

• Occurs consistently, every time
• Version: 2026.2.21-2
• Channel: Discord
• The content itself is harmless - it is the channel topic injected as context