What is accepted about one ref, and on what evidence

?target=<space>:<id> — for example memory:mem_123 or assertion:a_9.

edges are the edges BEARING ON this ref (those whose to names it). Edges leading FROM it — the case where this memory is a premise supporting something else — are not here; ask warrant-impact, or use the per-memory route which does both.

The edges arrive WITHOUT labels, so no per-edge verdict is available and this response groups them by the polarity each states. The headline status is the fold's verdict and accounts for every defeat.

An edge is a REASON, with an exact id. Two edges whose prose resembles are two edges and no fold merges them; the fold never reads reason at all, which is why it is required — it is written for the human reading the audit, and nothing else will ever supply it.

There is NO delete. Withdrawing a reason is bottom, an APPEND that leaves the edge, its defeat, and the history of both in the log forever.

An acceptance status is a joint function of TWO logs — the warrant log and the assertion log the verdict oracle reads — so every read carries both cursors. Two answers are comparable only at equal pairs.

contested is in the status vocabulary and CANNOT OCCUR in the current topology: an edge attacks exactly one target, so undercut cycles are closed under their own targets and a support/rebut edge can never be undecided. Do not build a UI state for it.

Requires the coord:read scope. Accepts all three lanes: a session JWT (bounded by the caller's role on this workspace) or an il_* machine credential whose own scope set, workspace set and validity window cover it — both ceilings apply and neither can raise the other, so a credential carrying coord:read is still refused where the role does not permit it.

A credential with no scope record resolves to LEGACY_DEFAULT, which does NOT include coord:read — mint a key with it explicitly (POST /v2/workspaces/{id}/keys).