anti-ai-writing

# Anti-AI Writing

> A Claude skill that edits any piece of writing to remove AI tells on two layers: the sentence level (banned words, banned phrases, banned structures like em dashes, the default rule of three, bulleted bold titles, formulaic transitions) and the discourse level (stated takeaways, single-track structure, manufactured closure, atmospheric openers, and other tells that survive a surface edit). Returns a clean rewrite plus a diagnosis, or runs a strict review that returns a verdict instead. Works in English and Spanish.

## What this repo is

A single skill, `anti-ai-writing`, packaged for two Claude environments:

- **Claude Code**: install as a plugin from this repo's marketplace.
- **Claude Cowork**: download the bundled `.skill` file from `dist/`.

The skill is a stylistic editor. It accepts a draft and returns a rewritten version with AI tells removed, plus a diagnosis explaining what changed and why.

## Primary skill files

- [skills/anti-ai-writing/SKILL.md](skills/anti-ai-writing/SKILL.md): entry point. Two-layer workflow (read, scan sentence + discourse layers, rewrite, reread aloud, output with diagnosis). Also defines strict review mode.
- [skills/anti-ai-writing/references/banned-list.md](skills/anti-ai-writing/references/banned-list.md): full Era 1 / Era 2 / Era 3 banned word list. Banned phrases. Banned structural patterns. Weak verb constructions. Spanish AI tells.
- [skills/anti-ai-writing/references/rewrites.md](skills/anti-ai-writing/references/rewrites.md): how to fix each sentence-level violation. Word substitutions. Phrase rewrites. Em dash rewrites. Rule-of-three rewrites. Bulleted-bold-title fixes.
- [skills/anti-ai-writing/references/discourse-tells.md](skills/anti-ai-writing/references/discourse-tells.md): the eight discourse-level tells that survive a surface edit, with how to spot each.
- [skills/anti-ai-writing/references/discourse-rewrites.md](skills/anti-ai-writing/references/discourse-rewrites.md): how to fix each discourse tell without flattening the writer's voice.
- [skills/anti-ai-writing/references/strict-review.md](skills/anti-ai-writing/references/strict-review.md): scoring rubric and output format for strict review mode (verdict + finding list instead of a rewrite).
- [skills/anti-ai-writing/references/examples.md](skills/anti-ai-writing/references/examples.md): before-and-after pairs across tweets, emails, LinkedIn, blog intros, marketing copy, About pages, internal comms, discourse-level rewrites, and Spanish examples.

## Install commands

Claude Code:
```
/plugin marketplace add avectats7/anti-ai-writing
/plugin install anti-ai-writing
```

Claude Cowork: download `dist/anti-ai-writing.skill` and open it in Cowork.

## When the skill triggers

The skill fires on edit requests where the goal is to remove AI fingerprints. Triggers: "make this sound less like AI," "remove the AI tells," "edit this so it doesn't sound like ChatGPT," "humanise this," "proofread for AI," "this reads like an LLM," "make this less corporate," "rewrite without AI words." Strict review mode fires on "strict review," "audit this," "score this," "does this pass," "review estricto," or any request for a judgement rather than an edit.

## Output format

Rewrite mode: every output has four blocks: Diagnosis (what was found), Rewrite (the fix), Changes (3-6 biggest edits with rationale), Notes (anything debatable). Strict review mode returns a verdict plus a finding list instead of a rewrite.

## License

MIT. Free to use, modify, and ship commercially.

## Sister skill

[copy-that-sells](https://github.com/avectats7/copy-that-sells) generates copy. This skill polishes copy. They pair.

## Maintainer

Tato Polanco. https://github.com/avectats7

## Optional

- [README.md](README.md): full project description for human readers.
- [CHANGELOG.md](CHANGELOG.md): version history.
- [CONTRIBUTING.md](CONTRIBUTING.md): how to contribute new banned items or examples.
- [docs/install.md](docs/install.md): detailed install guide.