List workspaces

A page of the active workspaces the caller can discover. Archived rows are excluded by default, and filter[workspace_status][eq]=archived (or [in]=active,archived) is the only way to see them — there is no separate flag, so a status filter governs the page outright instead of fighting a default. deleted rows are never listed, however you filter.

Supports the shared grammar: filter[field][op]=, sort=, q=, limit/offset, count=false. Fields: name, display_name, description, kind, workspace_status, parent_id, created_at, updated_at. The bare spellings workspace_status, status, kind, parent_id and name are refused with a 400 naming the filter[...][eq]= form, rather than silently widening the page as they used to; any other unrecognised parameter is still ignored.

include=member_count attaches each row's roster size from one grouped read. It is absent rather than zero on a workspace the caller can discover but not read into: this list runs under discover, which reaches through discoverable_by, while the count is taken under read on the membership rows — so a headcount is not something discovery alone reveals.

include=artifact_counts attaches each row's artifact_count and its per-kind artifact_count_by_kind breakdown, from one grouped read — the pair the pre-Postgres stack served here and the cutover dropped. Unlike member_count it is present on every row of the page, including a workspace the caller can discover but has not joined, and a 0 therefore means zero: it is taken under the page's own discover receipt, because a count of artifacts identifies nobody. It counts what POST /aggregate counts — live heads, no system artifacts, no archived ones, discoverable kinds only — so the number does not change when the caller joins.

include=permissions attaches each row's advisory can_* capability matrix. Advisory only — a UI gates controls on it; it never authorizes a request.