Are Decisions From This Session Saved?

One question, asked at the end of a session: is there anything in the planning of this system that would be lost if we stopped here? Whatever the session settled about the thing being built exists only in the transcript until someone writes it on a surface, and the next session doesn't read transcripts. This skill finds those, checks each against the surface it belongs on, and closes the gap — recorded, or dropped with the human saying so out loud. It ends by answering the question straight: no, nothing about the design would have to be decided again; you can close the session — or, if that isn't true yet, exactly what would.

The subject is the system, not the session. How the work got done here — which skill was called, how the human asked you to proceed, what you were corrected on mid-task — is not planning and is not in scope. See What is not in scope; raising that material buries the two decisions that actually matter under ten that don't.

The session is the source, not the tracker. Every other planning skill starts from an issue and works down. This one starts from what was decided here and asks whether any of it landed. A decision that exists only in this conversation is one context window from never having happened.

Nothing is written and nothing is dropped without the human. Every entry gets recorded where they say, or forgotten because they said to. That is why this skill is safe to call at the end of a session without asking first — it proposes; the human decides each one.

What counts

The test: would the system be planned or built differently by someone who never saw this conversation? If yes, it's in scope. Everything below follows from that one question.

• an answer to a question the map or a ticket was open on
• a choice about the system's shape, behaviour, or contracts — including "we tried X, we're not doing X"
• a constraint or boundary found mid-work — "this can't land before Y"
• a scope call — something ruled into or out of this effort