Device "turned_off" (also) always triggers on a reboot. How to avoid?

Configuration
1

I have an automation like this:

alias: Climate / IR Panels DiningRm / Off
description: ""
trigger:
  - platform: device
    type: turned_off
    device_id: 21ba47d6953bff8a6e79c485285d9391
    entity_id: e3b730ee7a3d580006cc7fde486eedd7
    domain: switch
  - platform: device
    type: turned_off
    device_id: 84f00fe3b7b122ffa7cec77dc6debf35
    entity_id: 5eff716f4d558fd6ae34608643a6a66c
    domain: switch
condition: []
action:
  - service: media_player.play_media
    target:
      entity_id: media_player.muse_luxe_diningrm
    data:
      media_content_id: >-
        media-source://tts/google_translate?message=Heating+Panels+off.&language=en-us
      media_content_type: provider
    metadata:
      title: Heating Panels off.
      thumbnail: https://brands.home-assistant.io/_/google_translate/logo.png
      media_class: app
      children_media_class: null
      navigateIds:
        - {}
        - media_content_type: app
          media_content_id: media-source://tts
        - media_content_type: provider
          media_content_id: >-
            media-source://tts/google_translate?message=Heating+Panels+off.&language=en-us
    enabled: true
mode: single

It’s supposed to play a message on a speaker (and other things which I have omitted here for simplicity) when the switch is turned OFF.

This works BUT the automation also triggers every time Home Assistant reboots.

How can I avoid this?

2

How/where is that switch you’re triggering on defined? A particular integration? Is this a template switch?

Show the history of the switch around the time of a restart. It might be changing state from unavailable, in which case you can use a state trigger with a not_from option. See my next note.

Also, rather use state triggers.

3

It’ll be going from “unavailable” or “unknown” to “off”. If it stays properly connected in normal use you could do this, using state triggers which are objectively better than device triggers:

trigger:
  - platform: state
    entity_id:
      - switch.ENTITY_ID_1
      - switch.ENTITY_ID_2
    from: "on"
    to: "off"

Alternatively / as well, you could install the uptime integration, and include a (template shorthand) condition:

condition:
  - "{{ (now() - states('sensor.uptime')|as_datetime).total_seconds() > 120 }}"

That will prevent the action being run within two minutes (120s) of an HA restart.