Human Writing Standards

Purpose

AI-generated text has recognizable patterns that erode trust and signal inauthenticity. This skill enforces rules for writing that reads as genuinely human—the standard used by Zed, Basecamp, Linear, Stripe, and Apple.

Workflow

To produce human-quality writing:

1. Draft content normally
2. Run the self-check (see Quick Self-Check below)
3. Fix any violations using the reference material

Quick Self-Check

Before publishing any content, verify:

1. Vocabulary scan: Search for banned words (delve, tapestry, landscape, leverage, robust, seamless, comprehensive, crucial, multifaceted, embark, journey, moreover, furthermore, utilize, foster, showcase)
2. Em-dash count: More than 2 on the page? Rewrite.
3. Opening check: Does it start with "In today's..." or "Whether you're..."? Rewrite.
4. Colon check: Does the headline have a colon? Remove it.
5. Vagueness check: Replace "helps you manage" with something specific.
6. Read aloud: Does it sound like a person talking, or a press release?
7. "The" count: Too many sentences starting with "The [product]..."? Vary it.
8. Enthusiasm check: More than one exclamation point? Cut some.

Core Rules (Summary)

Reference

For complete lists of banned words, replacement suggestions, and before/after examples, see:

• references/writing-rules.md — Full vocabulary lists, punctuation rules, and examples