Menu
Route voice input from the realtime voice agent pipeline based on message tags.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
MOVED — canonical skill now lives in the platform skills tree.
Self-enroll a voice for speaker recognition. Activate ONLY when (a) a mic `Unknown Speaker:` turn carries a clear self-introduction ("I'm X", "my name is X", "tôi là X"...), (b) prior same-tag `[voice:voice_N]` turns can be combined with a now-named turn, (c) a Telegram voice-note carries an intro, OR (d) the user asks to list / forget / identify voices. Do NOT activate on bare `Unknown Speaker:` turns with no name and no prior same-tag history — those need only a short ack. Self-enrollment only.
Proactive coaching across hydration, breaks, meals AND posture. Use when an [activity] event fires (message starts with `[activity] Activity detected: <labels>.` — labels include drink, break, or sedentary raw labels like "using computer"; sedentary events may also carry a [posture_summary: {...}] block when the user has been at the computer long enough for posture to drift), or when the user asks if they should drink water / take a break / fix their posture. Thresholds are computed from per-user logs, never guessed.
Push Claude Code activity to the user's device (e.g. a lamp) over the LAN. The plugin POSTs semantic events to a companion daemon on the device (HTTP, port 5002); the device turns them into its own feedback (LED / round display / voice). Supports connecting a device, pushing usage, sending notifications, and tuning how/when the user is alerted. Triggers: "connect my device", "push my usage to my device", "notify my device", "send to my device", "mute my device", "warn me earlier", "set warning threshold to N", "stop the task done notification".
Control the device's RGB light when the user asks for a SPECIFIC color (e.g. "yellow", "red", "yellow", "red", "turn on color X", "enable X light"), an LED effect, or turning LEDs off. Do NOT use for ambiance/activity lighting (use Scene) or emotion expression (use Emotion).
Camera control — snapshot, stream, and privacy toggle. Trigger on "what do you see", "look at this", "take a photo", "don't look", "stop looking", "stop watching", "stop staring", "camera off", "camera on", "give me privacy". MUST call [HW:/camera/disable:{}] or [HW:/camera/enable:{}] when toggling — never just reply with text.
| name | input-branching |
| description | Route voice input from the realtime voice agent pipeline based on message tags. |
Voice input passes through a realtime voice agent before reaching you. The agent handles chit-chat directly and delegates everything else. Message tags tell you which path was taken.
[voice-instruction] Play jazz music on Spotify
[transcript] play some jazz please on spotify
[voice-instruction] — the realtime agent's clean summary of what the user wants. Use this as the primary input.[transcript] — raw STT transcript. Often inaccurate — STT is locked to one language while the user may speak another. Treat as noisy supplementary context only, never as the source of truth.turn on the lights
No tags. Process as a normal voice event.
[HANDLED] Hey, how's it going?
[REPLY] I'm doing great! How about you?
[HANDLED] — the user's original message, already answered via TTS.[REPLY] — what the realtime agent said. The user already heard it.This is a history entry only. The conversation already happened. You are being notified, not asked to respond.
[voice-instruction] is the primary input. When present, use it over [transcript].[HANDLED] → always NO_REPLY. No exceptions. Even if the reply seems wrong or incomplete — the user already heard it. Do not correct, echo, paraphrase, or add to it.[HANDLED] silently. If the exchange reveals mood, intent, or information worth tracking (fatigue, stress, preferences), update memory/mood/wellbeing.[voice-instruction], [transcript], [HANDLED], [REPLY] are routing metadata, not user-facing text.Delegated — use instruction, ignore noisy transcript:
[voice-instruction] Set brightness to 50%
[transcript] set the brightness to like half
→ Process "Set brightness to 50%". Route to LED skill.
Delegated — transcript in wrong language (user spoke Vietnamese, STT is English):
[voice-instruction] Turn off the light and play some relaxing music
[transcript] ton of delay and play some relate music
→ Process the [voice-instruction]. The transcript is gibberish because STT tried to transcribe Vietnamese as English.
Delegated — transcript garbled by accent/noise:
[voice-instruction] What is the weather like tomorrow
[transcript] what is the wetter like to morrow
→ Process the [voice-instruction]. The transcript has misheard words.
Delegated without instruction:
play some jazz music
→ Process as-is. Route to music skill.
Handled — silent:
[HANDLED] What time is it?
[REPLY] It's 3:15 PM.
→ NO_REPLY.
Handled — log mood:
[HANDLED] I'm so tired today
[REPLY] That sounds rough. Maybe take a short break?
→ Log fatigue via mood/wellbeing. NO_REPLY.