Looking for a critical review of my 66 automations: presence, modes, appliances, and security on HA Green

## What I'm asking for

I'd really value a critical review of my Home Assistant automation setup. I want experienced eyes on it to spot:

  • Reliability issues
  • Race conditions
  • Missing safety guards
  • Duplicate or conflicting logic
  • Over-complex templates where native HA features would do
  • Startup edge cases
  • Unreliable triggers or bad time windows
  • Notification action collisions

I'm not looking for a full rewrite. I want targeted, prioritised feedback I can act on safely.

## Setup overview

  • Platform: Home Assistant Green
  • Cloud: Home Assistant Cloud (Nabu Casa) for Alexa
  • Automations: 66 (one file)
  • Scripts: 6 central scripts including activate_sleep_mode, activate_away_mode, activate_home_mode
  • Approach: Production-style, stability over cleverness

## Key integrations

  • Aqara FP2 presence sensor (9 zones, whole-home overall)
  • iCloud3 v3 (device tracking, Apple Watch)
  • ConnectLife (washing machine)
  • Tapo smart plugs (TV, microwave, fairy lights, etc.) with power sensing
  • Hildebrand Glow / Bright (electricity dashboard)
  • Blink (kitchen camera)
  • HASS.Agent (PC control)
  • Thames Water
  • Samsung TV
  • Alexa via Nabu Casa

## Architecture and philosophy

Mode-based design
Three central scripts handle Home, Away, and Sleep transitions. Automations call these rather than duplicating logic.

Presence detection
Combines several signals to reduce false positives:

  • iCloud3 GPS
  • WiFi SSID match
  • FP2 overall presence
  • Power consumption signals
  • A high_confidence_home boolean for stronger checks

Safety guards on device-controlling automations

  • input_boolean.automations_master_switch
  • input_boolean.startup_lockout (90 second post-restart lockout)
  • input_boolean.sleep_mode_active
  • person.jacob_lane: home on automations that must not fire for others entering the flat
  • Debounce for: timers on FP2 off-triggers to prevent flicker

Failsafes
Dedicated failsafe automations for washing machine and dishwasher to prevent plug shutoff during active cycles.

Audit logging
Optional structured JSONL audit log via shell command, gated by input_boolean.audit_logging_enabled. Categories include presence, modes, automation lifecycle, notifications, system log, HA lifecycle, critical entities.

Work-shift-aware planning
Template sensors derive bedtime, wake time, leave-for-work time from a Google Calendar plus user-configurable bedtime buckets.

## Known issues already on my list

I'd still welcome opinions on these, but they are tracked:

  1. Washing machine failsafe occasionally misfires when the plug is turned off manually with no active cycle. CSV logger is in place to map ConnectLife status values.
  2. Possible duplicate automations handling cycle completion and plug shutoff.
  3. Occasional Nabu Casa Alexa relink issue after entity registry changes.

## Where to look

GitHub repo with redacted YAML:

  • automations.yaml (66 automations)
  • scripts.yaml (6 scripts)
  • helpers.yaml (input_booleans)
  • entities.yaml (entity list with last states)
  • configuration.yaml (templates and shell commands)

## How to give feedback

Anything useful, from a one-line "this automation has a race condition" to a full review. I'd particularly appreciate:

  • Spotting where I'm using Jinja templates when a native HA condition or trigger would do
  • Spotting where for: debounce is missing or too short
  • Spotting startup edge cases
  • Spotting notification action key collisions
  • Spotting unnecessary mode: parallel or risky mode: queued setups
  • Spotting time windows that are effectively always true

## Background

I run this setup like a production system. Stability and predictability matter more to me than cleverness. I have ADHD and autism, so consistent, low-noise behaviour from the house matters for daily life.

Apologies if this is posted in the wrong place!

Thanks in advance to anyone willing to take a look.

For once, this is something that AI might be good at.

Not one of the chat things - if you look around the forum you'll find several ideas about systems that can read your config and comment on the kind of issues you mention.

I wouldn't suggest this normally, but judging by your post you already have a very good grasp of the detail. You should be well qualified to extract some useful answers using AI as a tool.

1 Like