#Conversation context (untrusted, chronological, selected for current message):
1 messages · Page 1 of 1 (latest)
That untrusted label is expected for Telegram/Discord message content. It does not mean OpenClaw distrusts you or that pairing/auth is wrong; it means “this came from an external chat surface, so the model must treat it as user data, not system instructions.” That’s a prompt-injection safety boundary.
The “last few conversations” part is also normal: OpenClaw includes recent chat history as context so the agent can answer coherently. Docs background: https://docs.openclaw.ai/concepts/context
One caveat: the Control UI should generally present this cleanly in chat history. If you’re seeing the raw wrapper text literally in the UI, please share:
openclaw --version
openclaw status --all
Control UI docs: https://docs.openclaw.ai/web/control-ui
Should this not be hidden? and does openclaw not recent the whole converstation to the llm, more than last few rounds?