How do I install the Ha Assist skill?

Run npx skillmds@latest add beclab/ha-assist in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Ha Assist safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Ha Assist?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Ha Assist free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Ha Assist?

beclab (@beclab) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Ha Assist by beclab — Agent Skill

Name: Ha Assist
Author: beclab

Ha Assist

Use Home Assistant's Assist voice/conversation layer with hass-cli: send a natural-language command to the conversation agent, list conversation agents and assist pipelines, list STT/TTS engines and voices, and speak text on a media player. Use for 'tell Assist to turn off the lights', 'what voice assistants are configured', 'run a sentence through HA' requests.

beclab f75c156 2.8 KB Updated Sep 17, 2026

File contents

ha-assist

Conversation, pipelines, STT/TTS. Prerequisite: ../ha-shared/SKILL.md.

Send a natural-language command

The conversation agent maps a sentence to actions — a quick way to control HA by intent instead of crafting a service call:

hass-cli raw ws conversation/process --data '{"text":"turn off the kitchen lights"}' -o json
# Optional: "language":"en", "agent_id":"<agent>", "conversation_id":"..."

Also available as REST: hass-cli raw api POST conversation/process --data '{"text":"..."}'.

Manage Assist pipelines

A pipeline chains conversation + STT + TTS (+ wake word). Manage them directly:

hass-cli assist pipeline list                 # all pipelines + preferred_pipeline
hass-cli assist pipeline get                   # the preferred one (--id for a specific)
hass-cli assist pipeline create --data '{"name":"My Assist","language":"en","conversation_engine":"conversation.home_assistant","conversation_language":"en","stt_engine":null,"stt_language":null,"tts_engine":null,"tts_language":null,"tts_voice":null,"wake_word_entity":null,"wake_word_id":null}'
hass-cli assist pipeline update <id> --data '{"name":"Renamed"}'
hass-cli assist pipeline set-preferred <id>
hass-cli assist pipeline delete <id>

hass-cli assist languages                      # languages supported by available pipelines
hass-cli assist devices                        # devices bound to a pipeline

conversation_engine is required on create; use conversation.home_assistant for the built-in agent. STT/TTS/wake fields can be null for a text-only pipeline.

Conversation agents (not pipelines):

hass-cli raw ws conversation/agent/list -o json

STT / TTS engines

hass-cli raw ws stt/engine/list -o json
hass-cli raw ws tts/engine/list -o json
hass-cli raw ws tts/engine/voices --data '{"engine_id":"tts.google_translate","language":"en"}' -o json

Speak text on a speaker

hass-cli service call tts.speak --data '{"entity_id":"tts.google_translate","media_player_entity_id":"media_player.kitchen","message":"Hello"}'

Notes

conversation/process is the highest-leverage entry point: one sentence, HA figures out the entities. Good for ambiguous requests.
Full speech (audio in/out) runs over assist_pipeline/run, which is a binary streaming flow not suited to a one-shot CLI; use conversation/process for text, and tts.speak for output.

---name: ha-assistdescription: Use Home Assistant's Assist voice/conversation layer with hass-cli: send a natural-language command to the conversation agent, list conversation agents and assist pipelines, list STT/TTS engines and voices, and speak text on a media player. Use for 'tell Assist to turn off the lights', 'what voice assistants are configured', 'run a sentence through HA' requests.---# ha-assistConversation, pipelines, STT/TTS. **Prerequisite:**[`../ha-shared/SKILL.md`](../ha-shared/SKILL.md).## Send a natural-language commandThe conversation agent maps a sentence to actions — a quick way to control HAby intent instead of crafting a `service call`:```bashhass-cli raw ws conversation/process --data '{"text":"turn off the kitchen lights"}' -o json# Optional: "language":"en", "agent_id":"<agent>", "conversation_id":"..."```Also available as REST: `hass-cli raw api POST conversation/process --data '{"text":"..."}'`.## Manage Assist pipelinesA pipeline chains conversation + STT + TTS (+ wake word). Manage them directly:```bashhass-cli assist pipeline list # all pipelines + preferred_pipelinehass-cli assist pipeline get # the preferred one (--id for a specific)hass-cli assist pipeline create --data '{"name":"My Assist","language":"en","conversation_engine":"conversation.home_assistant","conversation_language":"en","stt_engine":null,"stt_language":null,"tts_engine":null,"tts_language":null,"tts_voice":null,"wake_word_entity":null,"wake_word_id":null}'hass-cli assist pipeline update <id> --data '{"name":"Renamed"}'hass-cli assist pipeline set-preferred <id>hass-cli assist pipeline delete <id>hass-cli assist languages # languages supported by available pipelineshass-cli assist devices # devices bound to a pipeline````conversation_engine` is required on create; use `conversation.home_assistant`for the built-in agent. STT/TTS/wake fields can be null for a text-only pipeline.Conversation agents (not pipelines):```bashhass-cli raw ws conversation/agent/list -o json```## STT / TTS engines```bashhass-cli raw ws stt/engine/list -o jsonhass-cli raw ws tts/engine/list -o jsonhass-cli raw ws tts/engine/voices --data '{"engine_id":"tts.google_translate","language":"en"}' -o json```## Speak text on a speaker```bashhass-cli service call tts.speak --data '{"entity_id":"tts.google_translate","media_player_entity_id":"media_player.kitchen","message":"Hello"}'```## Notes- `conversation/process` is the highest-leverage entry point: one sentence, HA figures out the entities. Good for ambiguous requests.- Full speech (audio in/out) runs over `assist_pipeline/run`, which is a binary streaming flow not suited to a one-shot CLI; use `conversation/process` for text, and `tts.speak` for output.

Frequently asked questions