Skill: Device Health Check

Purpose

Perform a nightly device health survey at 4am. Query all connected Home Assistant entities for
online status, battery levels, and last communication time. Flag offline devices, critically low
batteries, and devices due for firmware updates. Store findings in memory and send an alert via
notify(channel="telegram", intent="send") — alert if issues found, all-clear summary otherwise.

When to Use

Use this skill when:

• The device-health-check scheduled task fires (cron: 0 4 * * *, nightly at 04:00)
• User requests "run the device health check" or "check device status"

Workflow

Step 1: Survey All Entities

Call ha_list_entities() (no domain filter) to get the complete list of entities.

Identify entities with problematic states:

• Offline/unavailable: state is "unavailable" or "unknown"
• Battery sensors: entity IDs or friendly names containing battery or battery_level
• Low battery: battery sensor value <= 20%

Build three categorized lists from the results:

1. offline_entities: entities with state unavailable or unknown
2. low_battery_entities: battery sensors with value <= 20%
3. critical_battery_entities: battery sensors with value <= 10%

Step 2: Get Battery Levels for Battery-Powered Devices

For each entity identified as a battery sensor (step 1), call:

ha_get_entity_state(entity_id=<battery_sensor_entity_id>)

Classify by urgency:

• Critical (<= 10%): "will fail within hours/days — alert required"
• High (11-20%): "replace within days"
• Medium (21-30%): "plan to replace within a week"

Step 3: Classify Findings by Severity

Classify each issue:

Condition	Severity	Action
Device offline > 24h (check last_changed)	Critical	Alert
Device offline < 24h	Warning	Store, include in next digest
Battery <= 10%	Critical	Alert
Battery 11-20%	Warning	Store, include in alert summary
Battery 21-30%	Info	Store for trend tracking
Entity state: unknown (transient)	Info	Store if recurring

Step 4: Store Findings in Memory

For each issue found, store a memory fact:

Critical/high battery:

memory_store_fact(
    subject=<device_friendly_name>,
    predicate="device_issue",
    content="battery at <X>% — needs replacement <urgency>",
    permanence="volatile",
    importance=8.0 if critical else 6.5,
    tags=["maintenance", "battery", "urgent" if critical else "warning"]
)

Offline device:

memory_store_fact(
    subject=<device_friendly_name>,
    predicate="device_issue",
    content="device offline — state is <unavailable/unknown> since <last_changed>",
    permanence="volatile",
    importance=7.5,
    tags=["offline", "maintenance", "urgent"]
)

Healthy check (no issues):

memory_store_fact(
    subject="device-fleet",
    predicate="device_issue",
    content="health check passed — all <N> devices online, no low battery",
    permanence="volatile",
    importance=3.0,
    tags=["health-check", "healthy"]
)

Step 5: Compose and Send Alert (if Critical Issues Found)

If any critical or high-severity issues were found, compose and send an alert:

Device Health Check — [Date]

Issues Requiring Attention:
  [device name] battery at 8% — replace soon (critical)
  [device name] offline since [time] — check power or connectivity

Warnings (non-urgent):
  [device name] battery at 15% — replace within a few days
  [device name] battery at 18% — plan to replace this week

[N] other devices checked — all healthy.

Send via:

notify(
    channel="telegram",
    intent="send",
    subject="Device Health Alert — [N] issues found",
    message=<formatted_alert>,
)

Use intent="send" — this is a scheduled proactive delivery, not a reply.

If no critical issues found, send a brief status update:

notify(
    channel="telegram",
    intent="send",
    subject="Device Health Check — All Clear",
    message="Nightly check complete. All [N] devices online. No low battery alerts.",
)

Exit Criteria

• ha_list_entities() called to survey all entities
• ha_get_entity_state() called for each battery sensor identified
• Findings classified by severity (critical, warning, info)
• memory_store_fact() called for each issue (and for a clean-bill-of-health if no issues)
• notify(channel="telegram", intent="send") called — either with alert (if issues) or all-clear summary
• Session exits — no interactive troubleshooting in this session (that is handled by the
  troubleshooting skill when user follows up)

Common Failure Modes

ha_list_entities() Returns Empty or Fails

• Store an error fact: memory_store_fact(subject="device-fleet", predicate="device_issue", content="health check failed — ha_list_entities returned no results", permanence="volatile", ...)
• Send alert: "Device health check could not complete — Home Assistant may be unavailable."

Too Many Devices to Check Individually

• Prioritize entities in unavailable or unknown state from ha_list_entities() (state is
  included in the listing response).
• Call ha_get_entity_state() only for battery sensors and offline devices — not for all entities.
• The listing result is sufficient to identify problems without individual state calls for healthy
  devices.

Transient Unavailability (Device Comes Back Quickly)

• If last_changed shows the entity returned to normal very recently (within 30 minutes), classify
  as Info rather than Warning.
• Still store the fact, but lower importance: importance=4.0.