# Receiving the audio handoff

The tutorial's button plays a real ggWave message, inspired by GibberLink. It does not speak the full tutorial or automatically configure another agent. A receiver needs a ggWave decoder. The bundled decoder accepts a downloaded or recorded mono PCM16 WAV at 48 kHz.

From this skill directory:

```sh
npm ci --ignore-scripts
npm run decode -- /absolute/path/to/pi-v1.wav --verify
```

No microphone is needed when decoding the downloaded WAV. For an actual speaker-to-microphone transfer, record with a compatible receiver such as the upstream ggWave/Waver demo. The recorded path has not been tested across physical devices; noise, volume and microphone processing can prevent reception. If necessary, convert a recording with `ffmpeg -i recording.wav -ar 48000 -ac 1 -c:a pcm_s16le converted.wav` before using this decoder.

The exact ASCII payload is `SKOA1|https://www.skoa.online/agents/pi-v1.jsonl|<64-character SHA-256>`. It uses ggWave 0.4.0, AUDIBLE_FAST, variable payload length. The checksum covers the exact UTF-8 JSONL bytes, including final newline. The page displays the same payload and provides the WAV as a download.

`--verify` downloads the JSONL from the allowlisted HTTPS origin and checks the full SHA-256. It prints metadata only; it does not execute any commands or fetch credentials. A checksum detects changed bytes, not the trustworthiness of instructions. Confirm the intended origin, read the runbook and follow the user's existing scope. The runbook points to the matching SKILL.md, bundle and pinned source example.

Sources: [ggWave](https://github.com/ggerganov/ggwave), [GibberLink](https://github.com/PennyroyalTea/gibberlink).
