telegram-voice-message

Files (2)

• SKILL.md
• openai.yaml · agents/

SKILL.md (first lines)

# Telegram Voice Message

Send a concise voice note to the topic that owns the current managed turn:

```bash
/usr/bin/python3 {{TELEGRAM_CONTROL_ROOT}}/agent_telegram.py voice --key completion <<'EOF'
The requested work is finished and verified.
EOF
```

Run this as a foreground task and keep the process attached until it exits.
Speech synthesis happens before the durable Telegram outbox entry is created,
so ending the turn while synthesis is still running can prevent the note from
being queued at all.

If the command runner yields a process or session ID, that is **not** delivery
confirmation. Resume or poll that same process until it exits and prints
`Telegram voice update queued.` (or an idempotent `already sent` result). Never
end the turn with that process still running.

Before telling the user that delivery succeeded, verify the durable outbox row
reached `sent` and that Telegram returned the current topic's thread ID. A
queued row, a shell session ID, or successful TTS generation alone is not proof
of Telegram delivery.

Choose a stable lowercase key describing this distinct message. Repeating the
same key and content is idempotent.

Use this only inside a Telegram Control-managed turn. Keep the message short
and never include secrets or unreviewed sensitive data. Voice text is submitted
to Microsoft Edge TTS. Continue to provide the normal final response.

From shantamg/telegram-control, Apache-2.0.

More from shantamg/telegram-control