Humanizer

Why it's ranked here

The package has a clear editorial method, broad installation options, and explicit safeguards against invented facts. Its 33-pattern checklist is unusually concrete. However, it remains a prompt executed by another agent, so the repository provides guidance and validation of packaging, not deterministic editing behavior.

What's good

Voice calibration studies sentence length, vocabulary, punctuation, recurring phrases, and transitions. A supplied sample can override generic style rules. The information-first policy permits restructuring without dropping claims, while the no-fabrication rule forbids adding unsupported facts, names, numbers, dates, quotes, or citations.

Tradeoffs

Several rules target conventions that may be intentional, including dashes, curly quotes, passive voice, title case, and common hyphenated pairs. A voice sample reduces that risk. Because the product is Markdown instructions, outcomes depend on the host agent interpreting and following them consistently.

How to use it well

Use it as a final prose-editing pass for essays, posts, technical writing, or reference material. Supply a representative sample when authorial voice matters, then inspect the rewrite for nuance. It does not research missing specifics, verify claims, or provide citations, so handle those needs separately.

Observed

License

MIT

Primary implementation

Plain Markdown agent instructions

Install surfaces

Skills CLI, Claude Code plugin, or manual skill-directory copy

Interface

Agent skill invoked by slash command, direct request, pasted text, or document target

Platform support

Any agent harness that supports skill-style Markdown instructions

Package validation

Dependency-free Python validator plus GitHub Actions checks