/vibesenpai:insights

Opens the insights pane beside the chat: your findings, the fixes you can apply, and the health of the history pass (which runs the free checks) and the model reads.

/vibesenpai:insights

It takes no arguments.

What it does

It shows what the history pass and the model reads have already found. It sends nothing to a model and spends no tokens; the model reads run only when a turn ends.

Claude Code prints Insights opened beside the chat. The pane stays open and refreshes when there is something new. A tour of the pane describes every part of it.

Where no pane can open

If Claude Code cannot place a pane, the command prints a text version instead:

1. the session you are in
2. your history
3. how far the model reads have got

From a demo run over the public test data of claude-code-log (MIT), after the model reads had finished:

  This session · 0 edits

  Your history

    11 edits were rejected and retried, across 7 files
       most often tokenizer.js
    3 misread intent
    5 inaccurate self report
    1 other

  The cache was rebuilt from scratch 30 times — 1.4M tokens written again

  Analysis

    101 of 101 sessions read
    Up to date.

  Run /vibesenpai:report for more details

The text version has no Apply and no answers. Those are only in the pane.

Related

/vibesenpai:judge	turns the model reads on or off
/vibesenpai:report	the same findings as an HTML file
Settings	which free checks run