What it is
A single-file agent skill that flags and rewrites "AI-isms" in a draft. MIT license. Packaged as a SKILL.md with platform frontmatter so it drops into any agentskills.io-compatible agent: Claude Code, OpenClaw, Hermes.
Close sibling to what Nic already runs in his own write-note flow, but more opinionated and portable.
Language
Markdown instructions. No code to run. The skill prompt is the product.
Install
Claude Code:
git clone https://github.com/conorbronsdon/avoid-ai-writing ~/.claude/skills/avoid-ai-writing
Or just copy SKILL.md into ~/.claude/skills/avoid-ai-writing/.
OpenClaw:
clawhub install avoid-ai-writing
Value
Two modes:
- Rewrite (default). Flags issues, rewrites the text, prints a changelog, runs a second-pass audit.
- Detect. Flags without rewriting. Trigger with "detect", "flag only", or "scan".
Audits against 36 pattern categories across content, language, structure, communication, and meta. Concrete examples:
| Pattern | Rewrite |
|---|---|
| "watershed moment" | concrete facts |
| "leverage" | "use" |
| "seamless" | "smooth" |
| "serves as" | "is" |
| uniform sentence rhythm | variation |
| generic conclusion | cut or replace with a claim |
Good reference for refining my own Nic voice rules. The copula avoidance and significance inflation lists are tighter than what I have in my-linkedin-style.json. Worth mining before my next guidelines pass.