geo-doctor
한 문장으로 넣는 GEO 도우미. 더우바오, DeepSeek, ChatGPT, Claude가 왜 나를 빼는지 찾고 고친다
Home Assistant skills for agents
스킬PythonGitHub @homeassistant-ai직접 설치운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
설치형 프로젝트는 배포 도메인 없이도 확인할 수 있습니다. 개인 저장소는 GitHub 로그인으로, 조직 저장소는 위의 저장소 확인 파일로 직접 확인해 주세요.
메이커가 쓴 소개는 위의 한 줄이 전부입니다. 더 자세한 내용은 저장소 README에서 읽을 수 있습니다. 저장소 README ↗
README 전체를 불러오는 중입니다.
README 전문 보기 ↗An Agent Skill is a portable Markdown knowledge pack that teaches AI coding agents best practices for a specific technology. This repository provides one for Home Assistant, following the open Agent Skills standard. Install it and your agent applies Home Assistant best practices in every session.
home-assistant-best-practices — a decision workflow and anti-pattern table in SKILL.md, backed by reference files the agent loads on demand. It covers:
Agents make a predictable set of mistakes with Home Assistant config — Jinja templates where a native trigger, condition, or helper exists, device_id where entity_id belongs, mode: single on a motion light, renames that silently break dashboards and scripts — and the anti-pattern table names each one with its fix. A typical entry:
# Without the skill: a template condition. HA checks only the Jinja syntax when the automation loads
condition: template
value_template: "{{ states('sensor.living_room_temperature') | float > 25 }}"
# With the skill: a native condition. HA checks every key, the entity ID format and the threshold when it loads
condition: numeric_state
entity_id: sensor.living_room_temperature
above: 25
See Skill Contents for the file-by-file map.
Requires Node.js 22.20+.
npx skills add homeassistant-ai/skills
Works with AI coding agents that support the Agent Skills standard: Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and 70+ others. It installs into the current project; add -g to install it for all projects. To update: npx skills update
Claude Code also accepts the installer above, which places the skill files in your .claude/skills/ directory. The commands below install it as a plugin instead, kept in Claude Code's own plugin cache.
Run each command separately inside Claude Code:
/plugin marketplace add homeassistant-ai/skills
/plugin install home-assistant-skills@home-assistant-skills
Run /reload-plugins or restart Claude Code for the skill to take effect.
Auto-update is off by default for third-party marketplaces. To turn it on, run /plugin, open Marketplaces, choose home-assistant-skills, and select Enable auto-update. To update by hand instead, run /plugin, open Marketplaces, choose home-assistant-skills, and select Update marketplace, then run /reload-plugins. From a shell, run these, then /reload-plugins or restart Claude Code:
claude plugin marketplace update home-assistant-skills
claude plugin update home-assistant-skills@home-assistant-skills
Both apps share the same Customize UI, reached via the left sidebar (Claude Desktop) or claude.ai/customize (browser).
homeassistant-ai/skills (or the full URL, https://github.com/homeassistant-ai/skills) → SyncThis installs the skill as a plugin synced from this repo — no download or zip needed. Sync automatically is on by default, so later updates arrive on their own. To pull an update at once, open Add → Manage marketplaces, then the marketplace's menu → Check for updates.
cd skills && zip -r home-assistant-best-practices.zip home-assistant-best-practices/Once installed, the skill loads on its own whenever the agent works on automations, scripts, scenes, helpers, dashboards, blueprints, or backups — it needs no explicit invocation. Try:
Create an automation that turns on the hallway light when motion is detected and turns it off five minutes after the motion stops.
The home-assistant-best-practices skill includes:
| File | Purpose |
|---|---|
SKILL.md | Decision workflow, anti-pattern table, and pointers to the reference files below |
references/safe-refactoring.md | Safe workflow for renaming entities and their display names, replacing helpers, restructuring automations; display-name override, config-entry, and storage-dashboard blind spots |
references/triggers-and-conditions.md | Purpose-specific and native triggers/conditions, for: durations, trigger IDs |
references/automation-actions.md | Waits, variables, automation modes, control flow (choose, repeat, parallel), error handling (continue_on_error, admin-only actions), disabling automations, measuring elapsed time via last_triggered, Mobile App notify action vs entity, trace retention |
references/helper-selection.md | Built-in helpers vs template sensors (with decision matrix) |
references/template-guidelines.md | When to use templates, when to avoid them, template sensor best practices, reusable custom_templates macros |
See CONTRIBUTING.md for guidelines on writing and submitting skills.
Skill misled your agent? Open an issue with the Report Bad Skill Advice template and let the agent fill it in — it holds the context to trace the failure to its source in the skill.
MIT
references/yaml-only-integrations.md| YAML-only integration types, post-edit actions (reload vs restart) |
references/device-control.md | Actions and targeting, entity_id vs device_id, buttons and remotes, domain-specific patterns |
references/scenes.md | Scene authoring: config shape, snapshot/restore, snapshot-vs-script distinction |
references/dashboard-guide.md | Dashboard layout, view types, strategies, sections, cards, badges, custom cards, CSS styling, HACS / Marketplace |
references/dashboard-cards.md | Built-in card type lookup (core, energy, legacy), where to read a card's fields, entities-card attribute rows |
references/domain-docs.md | Integration and domain documentation (actions, entity attributes); doc pages for specific triggers, conditions, and actions |
references/examples.yaml | Compound examples combining multiple best practices |
references/appdaemon.md | AppDaemon apps: when to use vs. native HA, app structure, actions, scheduling, error handling, safe refactoring impact |
references/blueprint-guide.md | Blueprint authoring: metadata & source_url, inputs & selectors, target vs entity, defaults, !input templating, versioning |
references/backups.md | Full instance backups vs rolling one object back, when to offer a backup before an operation, what an archive contains, encryption keys, restore verification, backup deletion, why a git config repo does not replace a full backup |