D321 — active means the page is valid, everywhere; waiting work is declared

Decision. The active status family is never open work, in a journal or a map. A journal declares what open means through open_statuses (map_create kind: journal writes [open, in-progress, blocked]); a KB that reads active as open lists it there. A concept declaring review_after today or later is not stale_open (still open, listed in Work, not stale), and waiting_on records who or what blocks it. status_semantics warns on write, and kb_review proposes the migration (status_reclassify) from deterministic signals.

Why. openPhase treated active as open in a journal only: a rule no author, agent or Work panel reader could see. On a ~760-concept KB the Work panel listed 92 "active" concepts, mostly assessments, registers and resolved incidents. The only escape for work blocked on a third party was touching timestamp, which hides its real age. The cost: a breaking change for KBs relying on the old reading, named in the release note.

Alternatives rejected. A grace period with a warning only: a warning without action is ignored. A per-KB opt-in to the new meaning: the confusion is the default, so the default must change. Silent migration by the server: it never decides content (D14), hence a review kind the doctor applies and the operator confirms. A new suspended_until field: review_after already means "come back after this date".

Consequences. openPhase and workPhase must keep agreeing, and reviewSuspended serves both stale_open and Work's stale. The review kind status_reclassify and the check status_semantics share activeNotOpen. A past review_after suspends nothing. Journal templates and the bundled kb-doctor skill must not prescribe active for work.