Give it a voice

Create consistent lifecycle audio for greeting, idle, time-limit, and error events.

Audio flow

[1] Write text ──▶ [2] Check credential ──▶ [3] Generate ──▶ [4] Validate

The voice in agent.toml must match the generated lifecycle audio.

Prompt your AI agent

Recommended

Use the `dino-ping` skill to create lifecycle audio for this agent.
Write short text for greeting, idle warning, time limit, and error events.
Keep the language, voice, organization identity, and tone consistent.
Check the stored provider credential without exposing it.
Show the generated roles and provider side effect before requesting approval.
Generate only after approval, then validate the complete agent.

Use the Dino CLI directly

dino credentials status --provider gemini

# After you approve provider usage and local file writes:
dino agent generate-audio --agent-dir <absolute-agent-dir> \
  --input <absolute-audio.json> --overwrite \
  --confirm-mutation agent.generateAudio

dino agent validate --agent-dir <absolute-agent-dir>

Gotchas

• Never paste a provider key into chat, a command argument, or the workspace .env.
• Import a key only from an owner-only file outside the repository.
• Provider generation can incur usage.
• The CLI requires ffmpeg with the libopus encoder.
• Regenerate only selected roles when the other recordings remain correct.

Next

Use Publish your agent after the full agent validates.