Status

Run patina status and interpret it. The numbers are not self-explanatory, and three of them are the ones that matter:

• Every review a no-op. The prompt is not reaching the model. This is the failure the reporting exists to catch, not a quiet library.
• Runs that returned no structured output. If that count is climbing, --json-schema is not being honoured and every review will read as "nothing to save" whether or not it found something.
• Skills written but never loaded. The loop is working and the library is not. An unused skill is almost always a description that never matched anything, rather than a body that was wrong — the fix is rewriting the trigger, not deleting the skill.

Also worth calling out when present: spend that is climbing faster than the number of reviews, a sweep backlog that is not shrinking, or runs stopped at the spend ceiling (retrying those cannot help — the ceiling is the problem).

Report what you see in a few lines. If everything is healthy, say that in one line rather than restating the whole table.