Decision Log

Decisions made in Slack threads, on calls, and in Claude chats evaporate. Three months later, "why did we go with Stripe?" gets four different answers and no source of truth. Worse — the same decision gets reopened repeatedly because nobody remembers it was already settled. A decision log creates the institutional memory most small teams and solo operators lack. This skill builds and maintains one.

Not every choice deserves an entry. The log is for decisions where reversing them would cost time or money, where multiple people need to know, where alternatives were genuinely debated, where future-you will forget the reasoning, or where the decision constrains future ones. Choosing React means not choosing Vue — that constraint is exactly the kind of fact that disappears unless written down.

Every entry has five components in a fixed structure. The Decision (one clear sentence — no ambiguity, two readers should reach the same interpretation). The Context (what triggered it). The Alternatives Considered (at least one, with the reason for rejection — one line each, not a full analysis). The Rationale (the two or three strongest reasons, plus any data that supported them). The Consequences (what this enables, what it constrains, what follow-up decisions it triggers). Plus a Revisit If condition — the specific trigger that should reopen the question.

Quick-log mode exists for fast capture. If the user provides enough detail upfront, the skill skips the interview and generates the entry directly, only asking follow-ups for genuinely missing critical fields. The point is to make logging frictionless enough that it actually happens — a log nobody updates is worse than no log at all.

The skill also runs staleness reviews. On request, it scans for decisions older than six months without revisit, decisions whose revisit conditions may now be met, decisions still marked Active on paused or completed projects, and contradictory entries that conflict. Decisions are superseded rather than deleted — the history matters, and a reversed decision tells you something about the trajectory of the project that a clean log doesn't.

The rationale field ages the best. "What" becomes obvious in hindsight; "why" is what stays useful for years. Writing for strangers — the version of you who's forgotten, the new hire who wasn't in the room, the cofounder who joined later — is the discipline that turns a private notebook into shared institutional memory.