Humanist — AI skill for Claude Code
A prose checker and an offline sanitizer for AI-assisted writing, with the blind studies behind both.
How to install Humanist
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open itsryanlenk/humanist and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Humanist does
A prose checker and an offline sanitizer for AI-assisted writing, with the blind studies behind both.
Alternatives in AI
- Skill Seekers Roadmap — Transform Skill Seekers into the easiest way to create Claude AI skills from any knowledge source - documentat 11.1k ★
- Simulink Agentic Toolkit — The Simulink Agentic Toolkit gives your AI agent both the tools and the expertise to work effectively with Sim 987 ★
- Fact Checker — Verify claims and facts with sources 653 ★
README
humanist
**Make an AI-assisted draft measurably better to read.** In a blind study, editors preferred the processed draft over the raw one **22 times out of 22**.
That is the claim, and it is the only one this repository makes. It does not make writing pass as human, it cannot defeat AI detection, and the studies in [docs/evals/](docs/evals/README.md) are the reason both of those sentences are here rather than in a footnote.
What it is
**The rewrite pass.** Siqi Chen's MIT-licensed [humanizer](https://github.com/blader/humanizer), vendored under `plugins/humanist/skills/humanist/third_party/`. It strips inflated significance, AI vocabulary, negative parallelisms, filler and fake-candid openers, rewriting rather than deleting. **This stage carries the measured benefit: 16 of 16 blind trials against the raw draft.**
**The checker.** `humanist.py`, a stdlib-only sweep over the trope inventory in `ai-tropes.md`. It reports FAIL and WARN counts, readability, anaphora runs, duplicate sentences and bold-bullet counts, and **it never edits**. Read the [honest note on this stage](#the-checker-stage-is-not-proven) before relying on it.
**The composition read.** Five tells stay manual forever, because no regular expression catches them without lying about its precision: dead metaphor, fractal summaries, analogy stacking, one-point dilution, near-verbatim repetition.
Install
/plugin marketplace add itsryanlenk/humanist
/plugin install humanist@humanist
Or clone it. Python 3.9+, standard library only, nothing to install.
git clone https://github.com/itsryanlenk/humanist
cd humanist
python scripts/validate_repo.py
Quickstart
cd plugins/humanist/skills/humanist
python humanist.py draft.md # the checker
python humanist.py draft.md --strip-quotes # drafts quoting primary sources
python humanist.py draft.md --lenient # internal docs: register rules become notes
python humanist.py --rules
Related Skills
Human Copywrite
A Claude Agent Skill that humanizes AI-written marketing and long-form copy: landing pages, sales pages, e-boo
Agent Tug Of War
Claude vs Codex tug-of-war debates: two model families argue, both judge blind, you get a decision record. CLI
Aegis Integrity
Open-source, offline, bias-aware academic integrity checker: plagiarism, AI-text detection, citation verificat
Content Writing
A structured content writing skill for Claude and ChatGPT that produces SEO- and LLM-optimized articles follow
Sloptrim
A local detector for AI-writing patterns. Scores every prose file your agent saves. Python standard library on
Slopbuster
Strips the AI tells out of prose, code, and academic writing. 152 patterns from real AI-vs-human analysis, two
Related Agents
Prose Checker
Provides craft-level feedback on prose quality. Use PROACTIVELY during bi-chapter reviews for line-level writi
Mod Offline Worker
Offline worker for a modding project, run by mod-orchestrator. It never touches the running game. It studies r
Ccxt PR Reviewer
End-to-end review of a CCXT pull request. Reads the diff, transpiles and builds in all six languages (TS/JS, P