ha-conversation-agent-augmentlisted

Augments an existing Home Assistant Custom Integration with one or more Voice & AI surfaces — intent handlers, a conversation agent, and/or LLM API tools — conforming to spec/ha/intents-conversation plus spec/ha/llm-api. Decides with the user which surfaces are in scope, then generates the registered intent handlers, the conversation platform entity, and/or the LLM tools and API registration, implementing domain-appropriate built-in intents, avoiding deprecated ones, and signalling tool errors as HomeAssistantError; the full class-and-signature contracts live in the skill body and specs. Activate on "add a conversation agent", "register intents", "expose tools to the assistant via the LLM API", or equivalent German requests. Do not activate for assist_satellite/stt/tts/wake_word entities (ha/entity-platforms-voice), registered services (ha-service-definition-add), greenfield scaffolding (ha-integration-scaffold), or deploying to a live HA instance.
nolte/claude-home-assistant · ★ 1 · AI & Automation · score 68

Install: claude install-skill nolte/claude-home-assistant

# HA Conversation Agent Augment Spec: `spec/claude/ha-conversation-agent-augment/en.md` (EN canonical) / `spec/claude/ha-conversation-agent-augment/de.md` (DE translation). ## Why this is a skill, not an agent - **Human-visible augmentation surface** — the user describes a Voice & AI capability and reads back the generated handlers/entity/tools and the conformance report; a skill keeps this on the visible command surface, like the sibling augment skills (`ha-config-flow-augment`, `ha-coordinator-add`, `ha-device-automation-add`). - **Scope decision up front** — which of the three surfaces (intents / conversation entity / LLM tools) and, for tools, expose-vs-consume, is a per-run dialogue the user approves before any generation; that belongs in the working context. - **Bounded, inline generation** — intent handlers, one `conversation.py`, and a small `llm.API` with its tools fit inline; no isolated agent context is needed. - Counter-dimension considered: the draft→validate loop could be an agent, but the surface-scope decision and the two-axis separation belong in the user's working context; skill wins. ## When this skill activates Use this skill to add **one or more** Voice & AI surfaces — intent handlers, a conversation agent, and/or LLM API tools — to an existing integration, so its devices respond to Assist sentences and/or it exposes tools to the assistant. ## When NOT to activate - Assist satellite / STT / TTS / wake-word **entities** → `ha/entity-platforms-voice`