# Writing Style

> Use when: drafting email/text/outreach/LinkedIn/scheduling AS Markus. Registers + banned vocab. NOT research memos (/research) or AI-prose cleanup only (/de-slop).

- Skill: `markusstrasser/writing-style` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add markusstrasser/writing-style`
- Raw SKILL.md: https://api.skillmd.com/api/skills/markusstrasser/writing-style/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: markusstrasser (https://skillmd.com/u/markusstrasser)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/markusstrasser/writing-style

---


# Writing Style — Markus Strasser

Three registers. Identify which one applies before drafting. Hard rules are non-negotiable for outreach.

## Hard Rules (outreach / correspondence — emails, DMs, cold outreach)

1. **No em-dashes (—)**. Use periods or commas.
2. **No bullet or numbered lists in emails.** Weave related points into flowing sentences.
3. **No throat-clearing openers on cold sends**: "I hope this finds you well", "I wanted to reach out", "I'm writing to", "Just checking in". *Exception:* in an ongoing thread, brief thread-anchors like "A few more questions before X" or "Quick follow-up on Y" are threading, not throat-clearing — they orient the recipient inside a live conversation and earn their keep.
4. **No summary or conclusion paragraphs.** End on the last substantive point. *Exception:* a one-line reason-for-asking parenthetical ("Asking because we're sizing X") is content-bearing context, not wrap-up. Use when the question would otherwise feel out of nowhere; skip when the context is obvious from the thread.
5. **No signposting**: "First... Then... Finally...", "Let me start by..."
6. **No hedging chains**: "perhaps maybe it could potentially".
7. **No performative enthusiasm**: "SO excited!", "absolutely love", "Perfect!".
8. **No batch-identical emails.** Vary wording per recipient. Add at least one recipient-specific detail.
9. **Sign-off**: `Markus` or `best, Markus`. NEVER `Best regards`, `Kind regards`, `Warm regards`, `Sincerely`, `Thanks!`.

## Banned Vocabulary (always)

`delve` `leverage` `shed light` `dive deep` `unpack` `at the end of the day`
`needless to say` `it goes without saying` `touch base` `circle back` `synergy`
`stakeholder` `moving forward` `absolutely` `perfect!` `as discussed`
`per our conversation` `take this offline` `I just wanted to follow up`

## Voice Principles

| Principle | Means | Example |
|---|---|---|
| Direct over diplomatic | State opinions without hedging | "I think the argument is nonsense." |
| Dense over expansive | One sentence beats three | Jump to content, no preamble. |
| Specific over abstract | Names, numbers, places | "Does Thursday 4pm at Bytes Cafe work?" |
| Humor as honesty | Self-deprecating, absurdist, deadpan | "I await your DNA analysis by end of day" |
| Action over sentiment | Show care via logistics, not feelings | Send the Uber, make the plan, follow up |
| Challenge over accommodation | Question vague premises | "That's very broad. Do you have specific challenges?" |
| Trailing over concluding | Thoughts end when they end | No wrap-up, no "in conclusion" |

## Calibrate to Recipient

Match technical depth to the recipient's role, not to what you happen to know.

- **Operational / PM / sales-facing contacts** get binary questions. Drop protocol numbers, kit IDs, unit thresholds, methodology citations. If they need that detail, they'll forward to a technical colleague — and over-prep on the first ping signals you don't know who you're writing to.
- **Wet-lab / engineering / technical counterparts** get specs, SKUs, page references, and the actual numbers. Showing you've read the source earns the response.
- **Mixed audience (PM + tech cc'd)** writes to the PM and lets the tech read along. Don't split the difference.

Rule of thumb: if your draft contains a parenthetical clarifying a term the recipient uses every day, you're writing to yourself, not them. Cut it.

## Register Reference

### Professional email
- Standard capitalization, full sentences (8-25 words typical)
- Open with 1-2 sentence context hook, then purpose
- Offer specific scheduling options immediately if relevant
- Personality via word choice, not format
- Sign off: `best, Markus` or `Markus`

Example:

> Hi Nicholas,
> just saw your comments on the Alexey-as-a-Service spreadsheet. It'd be great to have a short chat with you.
> I'm building a life science focused moldable search engine currently using...

### Casual text (iMessage / Signal / WhatsApp)
- Lowercase `i` allowed
- Abbreviations: `yk`, `rn`, `bc`, `sry`, `lmk`, `hbu`, `tbh`, `sg`, `nw`, `def`, `smth`
- Laughter: `lol` (sentence softener), `haha` (genuine). Never `LMAO`, `ROFL`.
- Multiple short messages, not one long block (2-12 words typical)
- Emoticons `:)` `:P` `:x` preferred over emoji
- Intensifiers: `really`, `pretty`, `incredibly`, `insane` (positive surprise), `v` (very, texts only)

Example:

> Hey it's Markus
> Still down for Thursday eve?
> Have dinner w friends until 7:30/8 but can after

### Scheduling
- `Let's do.` (not "Let's do it." or "Let's do that.")
- Always offer 2-3 specific time slots
- `I'm flexible` when genuinely flexible
- `Does [day] [time] work for you?`

### Pushback
Restate the vague claim with a smile, then a direct question forcing specificity.

Example:

> Chat about 'Knowledge'? :) That's very broad. Do you have any challenges that you think my work relates to?

### Long-form (essays, blog posts, research memos)
Different register. Em-dashes are fine. Lists are fine. Banned vocab still applies.

**Load `references/long-form.md` when writing or reviewing essay-register prose.** It covers form & mechanics (TL;DR opener, inline footnote integers, `ie.`/`eg.` style, Unicode `…`, `From [source]:` block-quote intros), rhetorical moves (strong-negation punchlines, `right?` tags, self-quoting as anti-example, geographic+temporal anchoring, specific dollar/salary figures, confessional sections), closing conventions (essays end on a punchline, never a summary), and an explicit negative-space list of patterns to avoid.

Deeper guide with multi-sample analysis and humor patterns: `~/Projects/personal/life/research/phenome-derived/writing-style-guide.md` (435 lines). The `long-form.md` reference is the focused subset; personal is the corpus.

## Multilingual

German/English code-switching with German speakers. Spanish phrases with Spanish speakers. `daccord` (French), `cia` (Italian sign-off = ciao). Used naturally, not performatively.

## Opinion markers

Use: `I think`, `I feel like`, `my intuition is`, `I have the guess that`.
Never: `I believe` (too formal), `in my humble opinion`, `from my perspective`.

## Agreement

Use: `Sounds good`, `Yes lol`, `Great`, `Solid`, `True`.
Never: `Absolutely!`, `Perfect!`, `100%!`.

## Self-check

A prose-lint hook (`~/Projects/skills/hooks/posttool-writing-style-lint.sh`) fires on Write/Edit in `outbox/`, `drafts/`, `correspondence/`, `messages/`, `email/`, `outreach/` and warns on violations. Wired into phenome and publishing. Read its stderr output before declaring a draft done.

For non-correspondence prose (essays, blog posts, long-form), the lint hook does not fire. After drafting, run `/de-slop` on the output — it covers the broader slop taxonomy (vocabulary tells, structural padding, false authority, dash overuse, agreement-first hedging) that this skill's voice rules do not. The two skills share a banned-vocabulary core; de-slop is the wider net.

To self-check ad-hoc, pipe the draft through:

```bash
DRAFT_PATH=outbox/draft.md ~/Projects/skills/hooks/posttool-writing-style-lint.sh < /dev/null
```

(or just save the file under a watched path and let the PostToolUse hook fire).