Outbound Voice

Outbound Voice

Every Composer draft must read like it was written by a [YOUR_FIRM] human, not by a model. This skill is the source of truth for that voice.

Files in this skill

• voice-dna.md — the extracted voice profile. Tone, rhythm, vocabulary, opens, closes, banned phrases. This is THE source of truth. Ships with PLACEHOLDER content — you MUST run the extraction prompt below before the Composer can draft.
• voice-dna-extract.md — the prompt that turns a corpus of past messages into voice-dna.md. Re-run quarterly or whenever reply rate drifts.
• message-templates.md — opener, follow-up, and backup frames. Use as starting structure only; rewrite every line in voice.

The extraction corpus (real past [YOUR_FIRM] messages used as input) lives in OneDrive, not in the skill. It contains PII and never enters the skill folder.

Quality bar

A draft passes the voice test if a reviewer cannot tell whether it was written by a [YOUR_FIRM] human last quarter or by the model today. If they can tell, add 10 more real messages to the corpus and re-run voice-dna-extract.md.

What the Composer does with this skill

1. Open voice-dna.md. Confirm it is not a placeholder. If it contains PLACEHOLDER, stop and tell the operator to extract first.
2. Pick the appropriate frame from message-templates.md:
   • first touch (no prior thread)
   • reply-warm (positive reply)
   • reply-objection (skeptical reply)
3. Rewrite the frame line-by-line so the result honors voice-dna.md.
4. Scan for banned phrases (see compliance/banned-phrases.md AND the banned-phrases section of voice-dna.md).
5. Save the draft (either to the prospect MD file's ## Draft (pending approval) section, or directly to Outlook Drafts via CreateDraftMessage — see outlook-ops/SKILL.md).

Structural complement: no-ai-slop

The bans below and voice-dna.md cover slop words. They do not cover slop structure. The shared no-ai-slop skill (cowork/no-ai-slop/) is the structural complement: it catches binary contrasts, colon reveals, faux-insight setups, fake-profound endings, and robotic rhythm. The Composer runs it as a final pass after the voice draft (see composer/SKILL.md). Use the two together: voice-dna.md defines how we sound, no-ai-slop removes how machines sound. They are not interchangeable.

Anti-patterns

• "I hope this finds you well"
• "I came across your profile and was impressed by..."
• "I wanted to reach out because..."
• "Quick question..."
• Any subject line with the word "Quick" or "Touching base"
• Any opener that does not name a specific signal from the dossier