A Claude Code skill that enforces Australian English, bans common AI-tell words, and keeps written output sounding like a real human wrote it.
- Australian English spelling throughout (-ise, -our, -ence)
- Bans 30+ AI-tell words like "leverage", "seamless", "delve", "ecosystem"
- Kills em dashes and other punctuation crutches
- Enforces natural tone with contractions, varied sentence length, and direct phrasing
- Self-check step scans output before delivery
Works on emails, Slack messages, blog posts, reports, presentations, and anything else a human will read. Particularly good at cleaning up text generated by other LLMs.
# Clone into your Claude Code skills directory
git clone https://github.com/AdenCJM/writing-style.git ~/.claude/skills/writing-styleThat's it. Claude Code picks it up automatically.
Before (AI-typical):
In today's rapidly evolving digital landscape, organisations must leverage cutting-edge technologies to seamlessly navigate the complexities of modern business. It is important to note that a holistic approach to digital transformation can yield extraordinary results, fostering synergy across the ecosystem.
After:
Businesses need to keep up with new technology or they'll fall behind. But the ones getting real results aren't just adopting tools in isolation. They're rethinking how their teams, processes, and systems work together.
The skill is a single SKILL.md file. Fork it and edit to suit your preferences:
- Swap Australian English for British English (minor changes)
- Add or remove banned words
- Adjust tone rules
- Change scope exclusions
Tested across 3 prompt types (email, blog post, AI text rewrite) with and without the skill:
| Metric | With Skill | Baseline |
|---|---|---|
| Pass rate | 95.5% | 63.7% |
| Token overhead | +9.3% | — |
The skill eliminates em dashes and AI-tell words reliably, with minimal overhead.
MIT