SCBE Voice Render Verification

Use this skill when the task is to make SCBE voice work checkable, reproducible, and canon-aligned.

Core Rules

1. Keep canonical tongue order KO, AV, RU, CA, UM, DR.
2. Preserve layer mapping L3 -> tongue mixture, L6 -> breath, L7 -> phase, L14 -> audio realization.
3. Emit or validate a sidecar packet whenever voice behavior is changed.
4. Prefer verifiable outputs: patched file, sample packet, validator result, targeted command output.
5. Do not claim breath realism from punctuation-only pauses.

Read Order

When this skill triggers:

1. Read references/source-map.md.
2. Read references/verification-checklist.md.
3. Read only the repo files touched by the request.

Do not bulk-load unrelated audio, story, or emulator material.

Workflow

1. Lock the task mode.

• Choose one: docs, packet-schema, renderer-integration, qa, or review.

2. Load canon.

• Read C:\Users\issda\SCBE-AETHERMOORE\docs\LANGUES_WEIGHTING_SYSTEM.md.
• Read C:\Users\issda\SCBE-AETHERMOORE\docs\specs\SCBE_VOICE_EMOTIONAL_TIMBRE_SYSTEM.md.
• Read C:\Users\issda\SCBE-AETHERMOORE\scripts\voice_gen_hf.py if code is touched.

3. Build or patch.

• For docs, keep the canonical layer mapping and fallback rules explicit.
• For code, emit voice_packet or line.wav.json sidecars instead of burying control state in prose.
• Preserve speaker_baseline separately from runtime timbre controls.

4. Verify.

• Run scripts/validate_voice_packet.py on a sample or generated packet.
• Run targeted syntax checks when code changes:
  • python -m py_compile C:\Users\issda\SCBE-AETHERMOORE\scripts\voice_gen_hf.py
• State whether the packet is minimal or layer14.

5. Report.

• Return changed files.
• Return validator status.
• Return what remains unimplemented in the renderer.

Packet Contract

Use one of two modes.

Minimal Packet

Use for preprocessing or planning before synthesis.

Required fields:

• speaker
• text
• profile
• scene_intensity
• tongue_logits
• speaker_baseline

Layer 14 Packet

Use for governed render output.

Required fields:

• speaker
• text
• tongue_mix
• timbre
• breath_plan
• phase
• render

Use assets/voice-packet.example.json as the baseline sample.

Verification

Default validator command:

python C:\Users\issda\.codex\skills\scbe-voice-render-verification\scripts\validate_voice_packet.py `
  C:\Users\issda\.codex\skills\scbe-voice-render-verification\assets\voice-packet.example.json

Validation should prove:

• canonical tongue keys are present
• tongue values are numeric
• tongue_mix sums to 1.0 within tolerance
• breath entries use allowed kinds
• timbre keys are complete

Resources

• references/source-map.md
• references/verification-checklist.md
• scripts/validate_voice_packet.py
• assets/voice-packet.example.json