Just got done setting up my first OpenClaw agent on a Pi. The docs kept saying "be careful with logs, they can leak secrets." But when I looked at my logs, I was confused.

The `tool call` log shows the *intent* (e.g., "send email to..."). The `tool result` log shows the *actual output* from the tool (e.g., the full API response with a key or an error message). Right now, both seem to use the same verbosity setting. If I turn down logging to hide a potential API key in a `tool result`, I also lose the useful context of the `tool call` that started it. Shouldn't these be separate controls? The call is usually safe, the result is often the risky part.

How are others handling this? Is there a config trick I'm missing, or is this a feature gap?