fix(orchestration): stop claiming a structured worker was checked for a prompt

worker-show reported observation.agentWait: null for every structured worker. The
field's own contract says null means Orca looked and found no wait, and absent means
it never looked — and nothing looks here: a structured worker parks on a journal
question item, which no terminal prompt scan can see.

So null was a false negative on the one field a coordinator is explicitly told to
read, and it was mode-dependent: the same worker as a PTY would have reported the
wait. Absent is both the honest value and a state a PTY worker already reaches (an
older host, an unreadable pane, a probe that did not answer), so it discloses
nothing about which mode ran.

src/main/runtime/rpc/methods/orchestration-worker-mode-opacity.test.ts

@@ -22,6 +22,7 @@ import {
 } from '../../structured-worker-identity'
 import { ORCHESTRATION_METHODS } from './orchestration'
 import { readStructuredWorkerOutput } from './orchestration-structured-worker-lifecycle'
+import { inspectWorkerTerminal } from './orchestration-worker-observation'
 
 const WORKTREE = 'repo::wt'
 const STRUCTURED_HANDLE = 'structworker_worker'
@@ -257,4 +258,24 @@ describe('a worker cannot tell which mode it is running in', () => {
     expect(read).toThrow(/--source auto or --source transcript/)
     expect(read).not.toThrow(/structured/i)
   })
+
+  it('never claims a structured worker was checked for a human-answerable prompt', async () => {
+    installStructuredCoordinator(STRUCTURED_HANDLE, 'sess_worker')
+    vi.spyOn(runtime, 'showTerminal').mockResolvedValue({
+      handle: 'term_coord',
+      worktreeId: WORKTREE,
+      status: 'running'
+    } as never)
+    const started = await startWorker({
+      settings: STRUCTURED_DEFAULT,
+      from: 'term_coord',
+      coordinatorPaneKey
+    })
+
+    const observation = await inspectWorkerTerminal(runtime, db, started.dispatchId)
+
+    // Absent, not null: null is the contract's "looked and found none", and a journal question is
+    // invisible to every prompt scan, so null would be a false negative a coordinator acts on.
+    expect('agentWait' in observation).toBe(false)
+  })
 })

src/main/runtime/rpc/methods/orchestration-worker-observation.ts

@@ -35,6 +35,11 @@ export async function inspectWorkerTerminal(
   if (structured) {
     // Exactness is the recorded pane and lineage, which the runtime getters answer from the
     // structured registry; there is no terminal to show.
+    //
+    // `agentWait` is deliberately ABSENT rather than null. Null is the contract's "Orca looked and
+    // found no wait", and nothing here looks: a structured worker parks on a journal question item,
+    // which no terminal prompt scan can see. Reporting null would tell a coordinator the worker is
+    // not waiting, which is the one thing the field's own documentation forbids inferring.
     const exact = db.isDispatchProcessCurrent({
       dispatchId,
       paneKey: structured.paneKey,
@@ -45,8 +50,7 @@ export async function inspectWorkerTerminal(
       terminal: null,
       exact,
       status: exact ? observation.status : 'identity_changed',
-      ...(exact && observation.reason ? { reason: observation.reason } : {}),
-      agentWait: null
+      ...(exact && observation.reason ? { reason: observation.reason } : {})
     }
   }
   const terminal = await runtime.showTerminal(terminalHandle).catch(() => null)