curating-decision-recordslisted

# Curating decision records A decision record exists to stop a decision being re-litigated. That makes the active set a working index, not an archive: every record in it is something a future reader is expected to read, so a record that no longer guides anything is not free — it is a tax on finding the ones that do. This skill owns retention and archival mechanics. It does not judge whether a recorded decision was correct — a record that documents a mistake is still worth keeping if it stops the mistake being remade. The goal is to reduce the active set **without erasing history that can still guide work.** Judge every record semantically. Read `records.path` and `records.required_sections` from `.ledger.yml`. ## Check supersession when you add a record Every new record triggers a scoped audit of active records covering the same decision, mechanism, or rejected alternative. Classify each full or partial supersession **while writing the new record** — not later. - Full supersession, already implemented: archive it in the same change. - Partial supersession, or independently useful rationale: keep it and cross-link. - An obsolete proposal: reject it, with an honest reason. - A rejection that no longer prevents a plausible mistake: delete it. **Do not defer a known match to a later corpus audit.** You are the only person who will ever have both records in view at once; a deferred match becomes two records that quietly disagree. ## Classify by remaining future value Age,