S13 · Notification Management
Declared state: built, not well structured.
This is the repository plan and evidence, not a live health reading. A task marked done may still depend on credentials, devices, deployment or a newer acceptance test.
Dependencies
No dependencies are declared in the catalog.
Tasks and gates
13.F1 · Push + Telegram delivery with quiet hours and suppression.
Tier F · done · Verification: hermetic-test
Gate: `test_proactive_suppression.py`, `test_proactive_windows.py`
- unverified-import · undated ·
docs/SYSTEMS-DETAILS.md:1970
13.F2 · No duplicate delivery — one sender, one path. *gate:* `test_acknowledgements.py`
Tier F · done · Verification: hermetic-test
Gate: `test_acknowledgements.py`
- unverified-import · undated ·
docs/SYSTEMS-DETAILS.md:1972
13.F3 · **Acknowledgement tracking.** Mushegh knows delivered vs seen vs acted-on, and an urgent
Tier F · done · Verification: planned-test
Gate: new `test_delivery_ledger.py` — an unacknowledged urgent item is re-raised once and only
- recorded-note · 2026-09-12 ·
docs/SYSTEMS-DETAILS.md:1973
13.R1 · Channel choice by urgency and context: voice if present, push if away, Telegram if
Tier R · done · Verification: hermetic-test
Gate: `test_notify_routing.py` [route] - voice if present, push if away, a silent Telegram text if asleep; each channel falls through to the next, 'suppressed' only when all are down.
- hermetic-green · 2026-10-17 ·
bench/test_notify_routing.py
13.R2 · Batching — non-urgent items accumulate into the next brief rather than interrupting.
Tier R · done · Verification: hermetic-test
Gate: `test_notify_routing.py` [batch] - a signal below the interrupt line (context_override, 0.85) that he is not there to hear waits for the next brief, spends no budget, is not repeated; the brief reads it once.
- hermetic-green · 2026-10-17 ·
bench/test_notify_routing.py
13.R3 · Escalation ladder with a stop condition, so nothing can loop.
Tier R · done · Verification: hermetic-test
Gate: `test_deadmans_switch.py` [5] + [6] (test_escalation_ladder) - off-box: one restart then one alert per outage; in-brain: repair on every red probe, the owner paged once, silence while red, recovery said once, the next outage re-arms.
- hermetic-green · 2026-10-17 ·
bench/test_deadmans_switch.py
13.E1 · A week with zero unwanted interruptions and zero missed urgent items, measured against
Tier E · open · Verification: manual
Gate: proactivity log review recorded in TODO.
HUD screen contracts
- Notification Management · /systems/S13
All subsystems · Source: docs/systems-plan.json.