refine

Refine — make the backlog ready (groom / enrich)

A refined backlog means the next person can pick the top card and start with zero questions. Two modes, same skill:

• Enrich (one card) — flesh out a single thin card.
• Groom (whole backlog) — a health pass over the top of the backlog.

Language / Idioma

Work and report in the user's language (EN/ES).

Enrich a single card

Bring a thin card up to "ready":

1. Context — what and why, in plain terms; link the relevant repo / second-brain notes.
2. Acceptance criteria — bulleted, testable ("returns 401 on bad password"). Pairs with quality/prove-it-validation.
3. Scope edges — what's explicitly out.
4. Size signal — rough effort; if it's clearly too big, hand to breakdown.
5. Write it back to the card description (--desc) and push if online.

Groom the whole backlog

A pass over the top ~10–15 cards:

1. Sync. python board.py status; if online, python board.py pull.
2. For each top card, judge readiness: has context + acceptance criteria + priority? If not, enrich it (above).
3. Re-order by priority/value so the top of the backlog is genuinely the next-best work.
4. Route the exceptions: too big → breakdown; dead/dupe → prune; unsorted inbox → triage.
5. Apply changes via board.py; push if online.
6. Report "Definition of Ready" status: how many top cards are ready vs. need work, and the ordered top 5.

Rules

• ⚠ Don't fabricate acceptance criteria or estimates — derive from the card/context, or ask. A confidently-wrong criterion is worse than a flagged gap.
• ⚠ Enrich, don't redesign — refinement clarifies intent, it doesn't quietly change scope. Surface scope changes to the user.
• ✅ "Ready" = clear goal + acceptance criteria + priority + right-sized. Anything missing one is not ready; say so.