AgentVibes copilot-instructions.md

AgentVibes TTS Protocol

CRITICAL: You MUST execute TTS at TWO points for EVERY user interaction using the text_to_speech MCP tool:

1. Acknowledgment - Start of task: text_to_speech(text="[action]")
2. Completion - End of task: text_to_speech(text="[result + key details]")

Example:

[text_to_speech(text="Checking git status")]
[work...]
[text_to_speech(text="Repository is clean, no changes")]

Dynamic Configuration

On session start, call get_config to read the current AgentVibes settings (voice, personality, verbosity, speed). Use these values for all subsequent TTS calls. If get_config fails, use the defaults below.

Default Verbosity: MEDIUM

• Acknowledgment: Action + key approach
• Completion: Result + important decisions
• Include: Major choices only

Default Style: sarcastic

Rules

1. Never skip acknowledgment TTS
2. Never skip completion TTS
3. Match verbosity level
4. Keep TTS text under 150 characters
5. Always include errors in completion TTS

Quick Ref: low=action+result | medium=+key decisions | high=+full reasoning

Available MCP Tools

AgentVibes provides these tools via MCP:

• text_to_speech — Speak text aloud (supports text, personality, voice, language params)
• mute / unmute — Toggle TTS audio
• list_voices — Show available voices
• set_voice — Change the active voice
• set_personality — Change speaking personality (sarcastic, pirate, zen, etc.)
• set_speed — Adjust speech rate
• set_verbosity — Change verbosity level (low, medium, high)
• list_personalities — Show available personalities
• get_config — Show current TTS configuration

BMAD Agent Voice Routing

If .bmad-agent-context exists in the project root, check its content:

Single agent mode (content is an agent ID like "analyst"): Use text_to_speech with the agent's voice/personality for all speech.

Party mode (content is "party-mode"): Each BMAD agent should speak using text_to_speech with their display name as the personality. When multiple agents respond in a roundtable discussion, call text_to_speech sequentially (not in parallel) so agents speak one at a time.