Clawmate Companion Tts

Workflow

1. Decide whether this reply should use voice.
2. If not, reply normally in text and do not call the tool.
3. If yes, write a short spoken version and call:

clawmate_generate_tts({
  text: "<short spoken text>"
})

Spoken Text Rules

• text is for speech synthesis, not for user display
• Keep it short, natural, intimate, and voice-note-like
• Usually 1-3 sentences
• Prefer spoken wording over written wording
• Do not use Markdown
• Do not use bullet lists
• Do not include code
• Do not turn a long explanation into spoken audio

Handling Tool Results

On success

1. Do not send the same content as visible text
2. If delivery.delivered is true, the plugin has already sent the voice message directly, so do not send it again
3. If delivery.delivered is not true, the tool returns a local audio path in audioPath
4. You must immediately use that path to send the voice message to the user
5. If nextAction = use_audio_path_to_send_voice is present, follow it literally and do not just repeat the path

On failure

1. Use the returned message to continue naturally in text
2. Do not fabricate audio paths or delivery syntax

Prohibited Actions

• Do not call TTS for code or long-form explanation
• Do not send duplicate visible text when voice is sent
• Do not invent local file paths
• Do not turn every reply into voice