hidden-thinking#

Thinking blocks show what a model is "thinking" --- it's inner monologue, which supports its next steps. These blocks are useful to get insight into how the model is performing.

Unfortunately, thinking blocks are hidden by default.

The scanner checks for this condition and provides steps to fix it.

To scan for hidden thinking blocks, run:

gage scan -s hidden-thinking -y

If the scanner detects hidden thinking blocks, it opens an issue that you and Claude can use to fix the issue. Use the /gage:review Claude command to kick off an issue review session.

#Manual checks

If you want to check a session directly, run gage session view and select a recent session. Navigate to a thinking entry and check the contents of the thinking block. If it's empty, thinking blocks are turned off (hidden).