Skip to main content
GET
Get what a judge is doing to a subject
Returns what a judge is doing to one subject now as data.activity. The newest judge dispatch on the subject (with judgeId, that judge’s newest) decides:
  • still running: kind: "running" with startedAt and its budgetMs. A dispatch older than twice its budget plus a minute is treated as dead, and the answer is null.
  • failed: kind: "failed" with finishedAt, errorKind and, when a governance refusal stopped it, the decisionId to look it up by. The failure’s raw text is never returned: it can quote model output.
  • otherwise activity: null, and the latest verdict is the answer.
subjectKind is one of skill, agent, procedure, workflow_graph or workflow_run, and subjectId is that subject’s stored key (see Subjects). Each parameter may appear at most once; a missing or invalid subjectKind, a missing subjectId or an empty judgeId is 400 bad_request. The read filters, so an unknown subject is empty, never 404. A PERSONAL agent, or one that no longer exists, is never a readable subject: for subjectKind=agent naming one, the answer is empty. Needs grants:read, policies:read and audit:read at the platform scope: a grant that covers only some org units receives 403, naming the first one missing.

Authorizations

Authorization
string
header
required

An External App: the JWT it obtains from POST /oauth/token with OAuth2 client_credentials. Pass as Authorization: Bearer <jwt>. An app's static companion key also authenticates as the app, but is deprecated and is not protected by Idempotency-Key.

Path Parameters

subjectId
string
required

The subject's stored key.

Minimum string length: 1

Query Parameters

subjectKind
enum<string>
required

The kind of subject.

Available options:
skill,
agent,
procedure,
workflow_graph,
workflow_run
judgeId
string

Only this judge's verdicts.

Minimum string length: 1

Response

The activity

data
object
required
meta
object
required