percept-voice-cmd
от jarvis563
Coding Agents & IDEsКратко
Voice command detection and action execution for OpenClaw agents.
Описание
Что делает этот навык
Detects wake words in ambient audio and routes spoken commands to an OpenClaw agent for execution. Supports built-in actions like email, reminders, calendar queries, and notes, plus arbitrary forwarding for custom commands. Speaker authorization limits command execution to approved voices.
Two-tier intent parsing (regex first, LLM fallback) keeps simple commands fast while still handling complex natural language.
Когда его использовать
- Sending an email hands-free while cooking
- Setting a reminder without touching a keyboard
- Checking today's calendar by speaking aloud
- Taking a quick note during a meeting
- Texting a contact while driving
Пример сценария
User says: "Hey Jarvis, remind me in 30 minutes to call the dentist"
- 1
Percept buffers incoming transcript and detects the wake word "Hey Jarvis"
- 2
Buffer extends 5 seconds to capture the full command
- 3
Intent parser matches the reminder pattern via regex
- 4
Contact and time parameters are extracted from the utterance
- 5
Command is dispatched to the OpenClaw CLI to schedule the reminder
Reminder set for 30 minutes from now with the note "call the dentist"
Требования
- percept-listen skill installed and running
- Percept dashboard accessible on port 8960 for wake word and speaker configuration
Установка
clawhub install jarvis563/percept-voice-cmdВыполните команду CLI или вставьте промпт в Claude Code, Cursor, Codex — любой кодинг-агент, которым пользуетесь.