calibrate-my-voice

To calibrate your customer support writing style, this Skill connects to your inbox or helpdesk via Composio, pulls 10–20 recent sent replies, and extracts tone cues like greetings, sentence rhythm, and sign-offs. It then writes a voice profile to config/voice.md for future drafts.

Yes, you can train your support voice using Gmail, Outlook, Intercom, Help Scout, or Zendesk. The Skill requires a connected inbox or helpdesk via Composio to pull your recent outbound customer replies and distill your natural phrasing and tone.

Tone analysis works by detecting greeting patterns, sentence rhythm, formality, sign-offs, and recurring phrases from your sent messages. It also identifies forbidden phrases to avoid, ensuring future customer support drafts match your actual writing voice.

The best way to fix tone drift is to re-run the voice calibration process. The Skill extracts fresh voice cues from your newest sent customer replies, updates the config/voice.md file, and logs the changes in config/context-ledger.json and outputs.json for traceability.

Yes, a connected inbox or helpdesk via Composio is required. The Skill needs to pull 10–20 real outbound customer reply samples to extract your repeatable voice cues and generate an accurate writing style profile.

You should run voice calibration after your support info setup is complete, specifically when the voice section is TBD. Re-running is recommended whenever you notice tone drift in your lifecycle communication or want to re-learn from newer sent replies.