environment-report - SKILL.md Agent Skill

name: environment-report description: Generate a daily home environment report from Home Assistant sensor readings.

Skill: Environment Report

Purpose

Generate and send a daily home environment report every morning at 8am. Check temperature, humidity, air quality, and lighting levels across all rooms, compare against the user's stored comfort preferences, and flag any out-of-range conditions with actionable recommendations. Deliver via notify(channel="telegram", intent="send").

When to Use

Use this skill when:

The environment-report scheduled task fires (cron: 0 8 * * *, daily at 08:00)
User requests "send me the morning home report" or similar

Workflow

Step 1: Discover Rooms and Sensors

Call ha_list_areas() to get all configured Home Assistant areas (rooms).
Call ha_list_entities(domain="sensor") to get all sensor entities.
Group sensors by area. For each area, identify relevant sensors:
- Temperature: entities with temperature in name/entity_id
- Humidity: entities with humidity in name/entity_id
- Air quality / CO2: entities with air_quality, co2, pm25, voc in name
- Illuminance / lighting: entities with illuminance, lux, light_level in name

Step 2: Read Current Conditions Per Room

For each room with sensors, call ha_get_entity_state() on each relevant sensor:

ha_get_entity_state(entity_id="sensor.living_room_temperature")
ha_get_entity_state(entity_id="sensor.living_room_humidity")
ha_get_entity_state(entity_id="sensor.bedroom_temperature")
# etc. for all rooms with sensors

Collect readings as a room-by-room map:

{
  "living_room": {"temperature": 72, "humidity": 48, "co2": 650},
  "bedroom": {"temperature": 68, "humidity": 52},
  "kitchen": {"temperature": 74, "humidity": 55, "air_quality": "good"},
  ...
}

Step 3: Retrieve Comfort Preferences from Memory

For each room, retrieve stored preferences:

memory_recall(subject=<room_name>, predicate="comfort_preference")

Also check for time-specific preferences (morning context):

memory_recall(subject="morning", predicate="comfort_preference")

If no preferences are stored for a room, use healthy default ranges:

Temperature: 68-76°F
Humidity: 30-60%
CO2: <1000 ppm
AQI: <50

Step 4: Compare Readings Against Preferences

For each room and metric, evaluate:

Within range: OK — note it briefly
Minor deviation (e.g., ±2°F, ±10% RH): Flag with soft suggestion
Moderate deviation (e.g., ±5°F, ±20% RH, CO2 1000-1500 ppm): Flag with recommendation
Critical condition (e.g., temperature <60°F or >85°F, CO2 >1500 ppm, AQI >100): Alert prominently

Step 5: Generate Actionable Recommendations

For each flagged condition, compose a concrete recommendation:

Low humidity: "Bedroom humidity at 25% — below your 30% minimum. Consider running the humidifier."
High CO2: "Kitchen CO2 at 1200 ppm — open a window or run the ventilation fan."
Temperature deviation: "Living room at 65°F — below your 68°F minimum. The thermostat may need adjustment."

Limit to 3 most important recommendations to avoid overwhelming the user.

Step 6: Store Deviation Facts in Memory

For any out-of-range conditions, store as volatile memory:

memory_store_fact(
    subject=<room_name>,
    predicate="comfort_deviation",
    content="<metric> at <value> — outside preference range of <min>-<max> at 8am",
    permanence="volatile",
    importance=6.0,
    tags=[<room_name>, <metric>, "morning-report", "deviation"]
)

Step 7: Compose and Send the Report

Format the report as a room-by-room summary:

Morning Home Report — [Day, Date]

Living Room: 72°F, 48% humidity — comfortable
Bedroom: 65°F, 52% humidity — temperature below your 68°F target
Kitchen: 74°F, 55% humidity, CO2: 1200 ppm — ventilation recommended
[Other rooms...]

Alerts:
  Bedroom is 3°F below your preferred 68°F. Want me to raise the thermostat?
  Kitchen CO2 at 1200 ppm — open a window or run the fan.

All other rooms are within your comfort range.

Send via:

notify(
    channel="telegram",
    intent="send",
    subject="Morning Home Report — [Day, Date]",
    message=<formatted_report>,
)

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

Exit Criteria

ha_list_areas() called to discover all rooms
ha_list_entities(domain="sensor") called to discover sensors
ha_get_entity_state() called for each relevant sensor per room
memory_recall() called to retrieve stored comfort preferences per room
Readings compared against preferences; deviations classified by severity
Deviation facts stored via memory_store_fact() for out-of-range conditions
Report composed and sent via notify(channel="telegram", intent="send")
Session exits — no interactive follow-up in this session

Common Failure Modes

No Sensors Available for a Room

Skip that room in the report (do not fabricate readings).
Note in the report: "[Room]: no sensors configured."

Sensor Returns Unavailable / Unknown State

Skip that metric for that room.
If it was working previously, store a volatile fact: memory_store_fact(subject=<room-sensor>, predicate="device_issue", content="sensor offline during morning report", permanence="volatile", ...)

No Comfort Preferences Stored

Use healthy default ranges and note at the end of the report: "Preferences not yet configured for all rooms. You can set room preferences by telling me what temperature, humidity, or lighting you prefer in each room."