decision-log has two modes. Capture extracts a decision from past context — a meeting summary, a Slack thread, a brief — and produces a structured decision record: decision statement, who decided, context, options considered, rationale, implications, and reversibility. Structure frames a pending decision: it defines the question precisely, maps the options and trade-offs, assesses reversibility, makes a recommendation if evidence permits, and names the decider and timeline.
Decisions are re-litigated because no one wrote them down, and when someone did write them down, the doc is unfindable six months later. decision-log solves both: it structures the record so it's useful when retrieved — not just what was decided but why, what was considered, and what would change the call. Every captured decision appends to a master DECISION-LOG.md in knowledge/decisions/, which means the full decision history is in one place.
Day 12 follows roadmap prioritization because the PM just made sequencing calls that should be captured while the reasoning is fresh. The skills in this week build on each other — the prioritization rationale (Day 11) informs the decision log (Day 12), which feeds the alignment memo (Day 13).