F5/F6 reminders: on the real cron path the platform refuses the inbox write (run has no organization), so scheduled reminders reach nobody until objectstack#16659 lands

[os-sam]

Blocked-by: objectstack-ai/objectstack#16659

Found while delivering #7 (card 12). credential_expiry_reminder (daily) and interview_reminder (hourly) are correct in what they select, mark and dedupe on both drivers and on the real cron path — but a run launched by the schedule trigger carries no organization, notify emits org-less, and this install (12 employer organizations + the platform one) refuses the tenant-scoped inbox/delivery writes under the platform's #8844 rule. Measured on cli 17.3.0 with the cron temporarily set to every minute:

• sqlite, two ticks: F5 selected=8 acted=4 then selected=4 acted=0; F6 acted=2 (a Quillstone and a Harborline interview, so the system sweep does span employers) then acted=0; reminder_sent / expiry_reminded_at stamped. A direct read of the sqlite file afterwards: sys_inbox_message EMPTY. Every recipient's GET /api/v1/notifications: empty. Boot log: [notify] no organization in scope per emit, plus Insert on 'sys_automation_run' was REFUSED: a system-context write on a tenant-scoped object must carry an organization ... holds 2 organizations.
• memory, two ticks: identical shape — same log lines, all recipient inboxes empty.
• The same two flows triggered through POST /api/v1/automation/<name>/trigger under a session (which threads that session's organization) deliver on both drivers: recipients read the rows back from their own inbox, and the second trigger inside the window sends nothing.

Nothing in this repo can thread the organization: NotifyConfigSchema has no organization key, the schedule and time-relative triggers set no tenantId, and a defineJob handler has no messaging handle at all. The flows will start delivering the moment the platform threads an organization for scheduled runs (or declares the user-keyed inbox objects platform-global) — no change here is expected beyond re-verifying.

Re-verification recipe (verification-only cron mutation, restore by blob hash afterwards): set both start-node crons to * * * * *, npx objectstack build, boot --fresh, create an ats_interview at now+24h, wait two ticks, read GET /api/v1/notifications as talent1@quillstone.example and candidate01@mail.example.

Related here: #39 (the tenant wall on ats_interview for platform personas — a different read-side symptom of the same posture).