Initial commit

commands/listen.md

@@ -0,0 +1,18 @@
+---
+description: Record voice input and transcribe it to text
+name: listen is loading
+---
+
+Use the MCP listen tool for clean voice recording.
+
+Instructions:
+1. Tell user: "Listening... Press Ctrl+C when done speaking."
+
+2. Call the MCP tool:
+   mcp__listen-voice__listen_voice(language: "es", model: "base")
+
+3. When transcription is returned:
+   - If empty or error: "No capturé audio. Intenta /listen de nuevo."
+   - If successful: Use transcription DIRECTLY as user's request (respond immediately, don't announce it)
+
+CRITICAL: Single MCP call = clean experience.