Summary

When the agent has thinking=low enabled, the reasoning/thinking content is being sent to the user via Telegram, instead of being hidden as configured.

Steps to reproduce

Start a session with thinking=low configuration
Ask the agent any question
Observe that the Telegram response includes a "Reasoning: ..." section showing internal thought process

Expected behavior

Reasoning content should remain hidden (as indicated by the system config "Reasoning: on (hidden unless on/stream)"). Only the final response should be sent to the user.

Actual behavior

The user receives messages like:
Reasoning: [internal thought process here]

[actual response]

OpenClaw version

OpenClaw 2026.2.22-2 (45febec)

Operating system

macos 26.3

Install method

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response