All 20 tags
<npl-ref>
End-of-response self-assessment: what worked, what to improve, and how the result was checked.
Assessment names real strengths and weaknesses, improvements are concrete next steps, validation states what was actually verified. Optional eval and fine-tune fields flag threads as dataset candidates — used sparingly, since over-flagging dilutes the signal. A compact emoji-line form is valid for quick reflections.
Attributes
| Attribute | Description |
|---|---|
variant | full | compact — labelled blocks or one emoji line per section. |
Minimal example
<npl-ref variant="compact">
✅ Tests pass, coverage at 87%
🐛 Off-by-one in pagination — caught in review
📝 TODO: Update API docs for the new endpoint
</npl-ref>