SessionStart hook cannot trigger an unprompted Claude response

SessionStart hooks can inject context into a new session, but there is no mechanism for that hook to cause Claude to respond automatically — the user must send a message first.

Context:
Using a SessionStart hook to restore actor/agent identity (e.g. reading a config file and injecting "you are Scribe-1, run /who and /hark"). After /clear or a fresh session, the hook fires correctly and context is injected, but Claude sits silently until the user sends any message.

Expected behavior:
If a SessionStart hook is configured (or a specific hook flag is set), Claude should automatically generate an opening response after the hook completes — e.g. running any /who+/hark style orientation commands and greeting the user.

Actual behavior:
Complete silence. The user must type something (even just "hello?") to trigger Claude's first response. There is no way to tell from the terminal whether the hook fired correctly until that prod is given.

Impact:
For agent/actor workflows with session identity, this creates friction and uncertainty on every session start. Users cannot confirm initialization worked without prompting. A hook-driven auto-greeting would close this gap cleanly.

---

Reported with assistance from Claude Sonnet 4.6 (Scribe-1 instance, Anthropic)