Claude Code Plugins

Community-maintained marketplace

Feedback

home-assistant

@cullenmcdermott/nix-config
1
41

This skill should be used when helping with Home Assistant setup, including creating automations, modifying dashboards, checking entity states, debugging automations, and managing the smart home configuration. Use this for queries about HA entities, YAML automation/dashboard generation, or troubleshooting HA issues.

Install Skill

Shared

Installs to .agents/skills, used by Codex, Amp, Warp, Cursor, OpenCode, and more.

CodexAmp
WarpCursorOpenCode
ClineGemini CLI
GitHub Copilot
Personal

Available across projects.

$npx skills-installer add @cullenmcdermott/nix-config/home-assistant --client shared
Project

Writes to .agents/skills.

$npx skills-installer add @cullenmcdermott/nix-config/home-assistant -p --client shared
Note: Review the skill instructions before using it.

SKILL.md

name home-assistant
description Work with the user's Home Assistant instance (ha.cullen.rocks) — look up entities, states, services, and config entries; write or fix automations, scripts, and Lovelace dashboards; and debug why an automation did or didn't run using its execution traces. Use for any Home Assistant question or YAML task, even when the user just names a device, room, or automation.

Home Assistant

Ground every answer in the live instance: look up the real entity IDs, service fields, and existing automations before writing YAML. Guessed entity IDs and service parameters are the most common way HA configs silently fail.

If Home Assistant MCP tools are available in the session, they cover the same queries — use whichever is present. Otherwise use the scripts below.

Scripts

In scripts/ next to this file (installed at ~/.claude/skills/home-assistant/scripts/). Run with uv run <path>; each declares its own dependencies. All need HA_TOKEN in the environment (a long-lived access token) and print JSON.

Script Args Use for
ha_search_similar_entities.py <pattern> Find entities by id or friendly name — start here
ha_get_entities.py [domain] List entities, optionally one domain
ha_get_state.py <entity_id> Current state + attributes
ha_get_services.py [domain] Services and their fields
ha_get_config_entries.py [domain] config_entry_ids for integrations
ha_get_config.py Version, location, loaded components
ha_get_automations.py [search] Automation configs — use similar ones as templates
ha_search_dashboards.py [pattern] Find dashboards (WebSocket API)
ha_trace_summary.py <automation_id> Success/failure rates, timings, common errors
ha_list_traces.py [automation_id] Recent runs with run IDs and status
ha_get_trace.py <automation_id> <run_id> Step-by-step trace of one run

Trace scripts take the automation's numeric id attribute (e.g. 1761430536701, from ha_get_automations.py or the entity's attributes), not the automation.xxx entity ID.

Writing YAML

  • Use current syntax: top-level triggers: / conditions: / actions:, each trigger keyed by trigger: state (not platform:), and action: for service calls (not service:). Match the style of the user's existing automations when they differ.
  • Always set alias, description, and an explicit mode.
  • Verify every entity ID and service field against the instance before presenting the YAML.
  • Telegram goes through telegram_bot.send_message with a config_entry_id in data (get it from ha_get_config_entries.py telegram_bot), not a notify.* service.

Applying changes

If you're working inside the HA config repo and it has its own instructions (CLAUDE.md, Makefile targets, validation hooks), follow those. Otherwise hand the user copy-paste YAML and say where it goes: Settings → Automations & Scenes → Create automation → ⋮ → Edit in YAML for automations; the dashboard's Raw configuration editor for Lovelace; configuration.yaml plus a restart for integrations.

Debugging an automation

  1. Find it and its numeric id: ha_get_automations.py <name>.
  2. ha_get_state.py automation.<name> — is it on, and when did it last trigger?
  3. ha_trace_summary.py <id>, then ha_list_traces.py <id> to pick a failing or suspicious run, then ha_get_trace.py <id> <run_id>.
  4. The trace shows which trigger fired, each condition's result, and where execution stopped. If it never triggered, check the trigger entities' actual states and attribute values with ha_get_state.py.
  5. Explain the cause from the trace evidence, then give the corrected YAML.