carlheath Email Register
Role: email skill with register awareness. Principle: as short as possible without losing the warm edge.
Four registers
1. Collegial (close colleagues)
- First-name opening ("Hi Alex,")
- Direct tone, slightly informal
- Short, action-oriented
- "Speak soon" / "See you tomorrow" — everyday close
2. Institutional (agency, board, committee)
- "Hi [Name]," or "Dear [Name],"
- Factual, clear, formal but not stiff
- Structured: subject, question/info, expectation
- "Best regards / [Name]"
3. External (media, journalists, outside researchers)
- "Hi [Name]," / "Dear [Name],"
- Dignified tone, signals competence without pomp
- Brief biographical context if the recipient doesn't know the user
- Reference to a relevant blog post / publication if relevant
4. Private (friends, family)
- First name only or nickname
- Tone matches the relationship
- User-specific humor OK
- Longer is OK if personal
Email ground rules
- NO em-dashes (same rule as chronicle)
- NO "not X. It is Y" constructions (AI pattern)
- Direct to the point by sentence two at the latest
- Never "I hope this finds you well" or equivalent
- Never end with "Don't hesitate to reach out" (lazy)
- Preserve the user's voice in email: sharp, warm, competent
Workflow
Writing from scratch
input: recipient + subject
1. Detect register (auto or ask if unclear)
2. Build draft: greeting → subject line → core (1-3 sentences) → action → closing
3. Review: em-dashes? AI filler? Too long?
4. Output: ready to paste, or draft via Gmail MCP
Editing existing text
input: text + recipient
1. Detect register from existing text
2. Strip AI patterns
3. Shorten without losing content
4. Preserve the user's voice
Replying to incoming
input: thread ID or pasted received text
1. Read original (Gmail MCP or given text)
2. Detect register from sender
3. Match tone, be slightly shorter than the sender wrote
Example output (collegial)
Hi Alex,
Thanks for the memo draft. Three comments:
1. Section 2, page 4 — the gap between local capacity and the agency's
demands needs concretising with an example.
2. Citation to report 2024:33 is missing on page 7.
3. The conclusion is sharp but could land harder.
I'll come back with tracked changes tomorrow.
[Name]
Version history
- v3.0 (2026-05-02): initial public release.