rabbit-writeslisted

# Rabbit writes Make writing read like a particular person wrote it. Two layers, and the order matters: 1. **The voice profile** is the person. It decides how the prose sounds and which rules are absolute. 2. **The engine** is everything true of good writing regardless of who is writing. It fills every gap the profile leaves. It lives in `references/` and `scripts/` beside this file. The profile wins every conflict. The guardrails below are the one exception, and they run the other way. A rewrite that clears every flag and reads sterile has failed. So has one that scrubs a real writer's habits into house style. **Paths.** `${CLAUDE_PLUGIN_ROOT}/skills/` below means the directory holding this skill and its siblings (`rabbit-writes`, `voice-setup`, `rabbit-readme-improver`, `rabbit-reads`, `rabbit-rewrites`, `rabbit-claude-md`). Claude Code expands the variable. On a host that doesn't, such as Codex, resolve it that way by hand. ## The override > Break any rule in this skill sooner than write something worse. Orwell's sixth rule outranks everything below. If a flagged word is the right word, keep it. If a rule would make a sentence clumsy, false, or less precise, the rule loses. ## What this skill claims These patterns are more common in machine text. They are not proof of anything. Detector audits find false-positive rates above 60% on non-native English writers (Liang et al., Stanford, *Patterns*, 2023). Open-source detectors show misclassification above 70% (Jabari