fix(subagent): persist descriptor before admission

packages/subagent/subagent-inprocess/README.md

@@ -12,7 +12,7 @@ The driver follows this sequence:
 
 1. Validate the parent depth and optional absolute `maxDepth`, then derive child depth as parent depth plus one and persist it in the child session header.
 2. Call `parent.ctx.agents.create` directly, passing the required request signal into the factory's creation transaction. A continuable request publishes exactly `request.continuation.sessionId` instead of an internally minted id.
-3. During that transaction's unpublished setup window, install the requested persona, tool restriction, structured-output runtime, and — for a continuable request — the one-shot `agent/step` contribution that appends the `subagent/descriptor` event after the initial `turn/start` and before the first request, so the descriptor reaches persistence with that turn's flush.
+3. During that transaction's unpublished setup window, install the requested persona, tool restriction, structured-output runtime, and — for a continuable request — the prepended one-shot `agent/prompt-submit` contribution. It appends the `subagent/descriptor` event before downstream prompt admission can block or throw; allowed admission opens the initial turn afterward, while the final required checkpoint persists the descriptor even when no turn opens.
 4. Publish the child, retain the returned `AgentHandle`, and drive one task with `child.followup(prompt)` followed by `child.whenIdle()`.
 5. For a continuable start or resume, call `child.ctx.sessions.flushRequired(child.session)` again before returning the result. This final confirmation requires an installed durability listener and retries events retained after a failed turn checkpoint; if no listener participates or any listener fails, `result` rejects with `SubagentError.code === 'DURABILITY_FAILED'`, retains the checkpoint failure as `cause`, and names the resumability risk in its message. Activation cancellation during this await owns the unpublished result even when the completed turn was already recorded or the checkpoint subsequently fails. Foreground runs keep the loop's best-effort checkpoint behavior.
 6. Read the child's own last assistant message and latest message-triggered turn reason, excluding any fork seed and later plugin-owned between-turn records.