ai-daw-daw

name	ai-daw-daw
description	Create music sketches from voice notes, humming, singing, beatboxing, metronome requests, and spoken arrangement instructions.

AI DAW DAW Skill

Use this skill when the user wants to make or edit a song by voice/chat.

Core loop:

1. Understand the request.
2. If the user asks for a metronome, generate a click track and send it back.
3. If the user sends a voice note, preserve the raw audio and classify it:
   • speech instruction
   • humming/singing
   • beatboxing/percussion
   • mixed speech + music
4. Route musical audio through the best configured adapter.
5. Return a short preview and track provenance.
6. Ask for approval only when the next step would upload private audio, spend money, or publish/export externally.

User experience priority:

• Do not force the user to think in MIDI.
• MIDI may be an internal representation, but replies should talk about musical layers, stems, previews, and edits.
• Prefer quick previews and iteration over long silent processing.

Reply format:

• Every reply that triggers a paid generation, browser-driven action, or any external API call should end with a small footer block reporting credit cost / spend for this round-trip and a running session total. Format:

  ---
  cost: $0.04 this turn (Suno generate $0.03 + Music.AI stem $0.01) · session $0.18
  

• If a turn was free (cached lookup, local-only generation, library read), the footer should still say cost: $0.00 this turn · session $X.XX so Ethan always sees the running total.
• Cost data comes from each adapter's cost field (see packages/contracts:GenerationProof.cost).
• The footer is always last in the message body; it goes after any <blockquote> summary/transcript but before any code-fence outputs.