Skip to content
Back to search

humanizer

MarketplaceWorks with Claude CodeGitHubIncomplete

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/aboudjem-humanizer-skill/badge.svg)](/listings/aboudjem-humanizer-skill)

HTML

<a href="/listings/aboudjem-humanizer-skill"><img src="/api/listings/aboudjem-humanizer-skill/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

**Recommended: the Vercel skills CLI.** `npx skills` is Vercel's open installer for agent skills.

npx skills add Aboudjem/humanizer-skill -g

Works in more than 70 agents: pick yours in the menu ([supported agents](https://github.com/vercel-labs/skills#supported-agents)). It installs for all your projects.

In a terminal, the command shows the agent menu. Without an interactive terminal, it installs into every agent folder, so use `-a <agent>` in scripts.

<a href="https://skills.sh/aboudjem/humanizer-skill"><img src="https://img.shields.io/badge/skills.sh-000000?style=flat-square&logo=vercel&logoColor=white" alt="skills.sh" height="24"></a> Humanizer's page: [skills.sh/aboudjem/humanizer-skill](https://skills.sh/aboudjem/humanizer-skill).

**Or install straight into one agent.** Each command installs for all your projects.

<p align="center"> </p>

<a id="install-codex"></a>**Codex**, tested:

npx skills add Aboudjem/humanizer-skill -g -a codex

<a id="install-claude-code"></a>**Claude Code**, install tested:

npx skills add Aboudjem/humanizer-skill -g -a claude-code

<a id="install-cursor"></a>**Cursor**, install tested:

npx skills add Aboudjem/humanizer-skill -g -a cursor

<a id="install-gemini-cli"></a>**Gemini CLI**, install tested:

npx skills add Aboudjem/humanizer-skill -g -a gemini-cli

<a id="install-github-copilot"></a>**GitHub Copilot**, install tested:

npx skills add Aboudjem/humanizer-skill -g -a github-copilot

<a id="install-opencode"></a>**OpenCode**, install tested:

npx skills add Aboudjem/humanizer-skill -g -a opencode

**Claude Code plugin.** The same skill, from the 10x marketplace. It adds no command or hook:

claude plugin marketplace add Aboudjem/10x && claude plugin install humanizer@10x

Only this project: drop `-g`. Another agent: find its `-a` code in [docs/editors.md](docs/editors.md).

<details> <summary>Install without an installer (curl)</summary>

This copies the one file the skill needs into your project:

mkdir -p .claude/skills/humanizer && curl -sL https://raw.githubusercontent.com/Aboudjem/humanizer-skill/main/skills/humanizer/SKILL.md -o .claude/skills/humanizer/SKILL.md

Swap `.claude/skills/` for the folder your agent reads. Use `~/.claude/skills/` for a global install.

</details>

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines
  1. {
  2. "name": "humanizer",
  3. "owner": {
  4. "name": "Adam Boudjemaa",
  5. "url": "https://github.com/Aboudjem"
  6. },
  7. "metadata": {
  8. "description": "See which words make a draft sound like AI, then rewrite it in one of five voices. Optional 0 to 100 score. No network calls of its own."
  9. },
  10. "plugins": [
  11. {
  12. "name": "humanizer",
  13. "source": "./"
  14. }
  15. ]
  16. }
README.md · README · 345 lines
  1. <picture>
  2. <source media="(prefers-color-scheme: dark)" srcset=".github/assets/hero-dark.svg">
  3. <source media="(prefers-color-scheme: light)" srcset=".github/assets/hero-light.svg">
  4. <img alt="humanizer: See why it sounds like AI. Two terminal cards: Before scores 57 out of 100 with stock phrases underlined, After scores 6. Fact check OK, all 7 facts kept." src=".github/assets/hero-light.svg" width="100%">
  5. </picture>
  6. <p align="center">
  7. <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-7C3AED?style=flat-square" alt="License: MIT"></a>
  8. <a href="https://github.com/Aboudjem/humanizer-skill/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/Aboudjem/humanizer-skill/ci.yml?branch=main&style=flat-square&label=CI" alt="CI status"></a>
  9. <a href="skills/humanizer/SKILL.md"><img src="https://img.shields.io/badge/patterns-55-FF006E?style=flat-square" alt="55 AI writing patterns"></a>
  10. </p>
  11. <p align="center">
  12. <b>English</b> · <a href="READMEs/zh-CN.md">简体中文</a> · <a href="READMEs/ja.md">日本語</a> · <a href="READMEs/es.md">Español</a> · <a href="READMEs/fr.md">Français</a>
  13. </p>
  14. **See why your draft sounds like AI, then fix it.**
  15. Humanizer quotes the words that sound machine-written, then rewrites your draft.
  16. | Know which words | Reads more human | Facts kept |
  17. |:--|:--|:--|
  18. | It quotes each word that sounds like AI. The CLI ranks them in a "Fix first" list. | The bundled example drops from 57 to 6 out of 100. Lower is more human. | The CLI's `compare --check-facts` fails a rewrite that lost a number, date, version or URL. |
  19. **Works with:** any agent that reads skills. It is not tied to one model: the skill is one Markdown file, and your agent's model runs it.
  20. Tested: Codex CLI. Install tested: Claude Code, Cursor, Gemini CLI, GitHub Copilot, OpenCode.
  21. Installs for all your projects:
  22. ```bash
  23. npx skills add Aboudjem/humanizer-skill -g
  24. # then, in your agent: /humanizer --file draft.md --voice casual
  25. ```
  26. [The example](examples/blog-post/README.md) that ships with the repo, before and after, scored by the CLI:
  27. ```text
  28. Before 57/100 Mixed
  29. In today's rapidly evolving technological landscape, artificial intelligence
  30. is reshaping how we think about creativity and artistic expression.
  31. After 6/100 Pristine
  32. Stable Diffusion 1.4 was released on 22 August 2022.
  33. Fact check: OK. All 7 fact(s) in the original survive the rewrite.
  34. ```
  35. Lower is more human. [Proof](#proof) names the command that prints these numbers.
  36. **What it is:** a skill that finds the words that sound like AI in your text, and rewrites them.
  37. **When to use it:** before you publish a README, blog post or email that you wrote with AI help.
  38. **What you get:** the rewrite in chat, in one of five voices. Your file changes only with `--mode edit`.
  39. It is not an AI detector, and it does not promise to pass one.
  40. ## What you get
  41. A report that starts with the verdict, then says what to fix. This is the optional CLI on the example:
  42. ```bash
  43. node cli/index.js score examples/blog-post/before.md
  44. ```
  45. ```text
  46. examples/blog-post/before.md
  47. Score: 57/100 (Mixed)
  48. Fix first:
  49. 1. Cut the stock AI words: 19 found ("it is worth noting", "it is important to note", "in today's", "rapidly evolving", ...). Costs 40.0 points (lexical).
  50. 2. Vary the sentence lengths: they run from 13 to 37 words (mean 24.4). Split a long one and add a short one. Costs 16.8 points (burstiness).
  51. 3. Rewrite it with the skill (/humanizer --file examples/blog-post/before.md), then run "compare --check-facts" on the pair.
  52. Detail:
  53. Signal breakdown (points out of 100):
  54. lexical 40.0 weight 0.4 vocabTellDensity 0.0697
  55. burstiness 16.8 weight 0.28 burstiness 0.263
  56. repetition 0.3 weight 0.14 trigramRepetition 0.004
  57. diversity 0.0 weight 0.18 mattr 0.843
  58. ... then 12 metric lines
  59. ```
  60. - **A 0 to 100 score** with one of five verdicts, from Pristine to Pure AI smell.
  61. - **A "Fix first" list** that names the words and sentences that cost the points.
  62. - **A rewrite in one of five voices**: `casual`, `professional`, `technical`, `warm`, `blunt`.
  63. - **A fact check** that fails a rewrite which lost a number, date, version, URL or acronym.
  64. - **An exit code** for CI, and a [pre-commit hook](docs/pre-commit.md) that scores only the files you staged.
  65. ## Install
  66. **Recommended: the Vercel skills CLI.** `npx skills` is Vercel's open installer for agent skills.
  67. ```bash
  68. npx skills add Aboudjem/humanizer-skill -g
  69. ```
  70. Works in more than 70 agents: pick yours in the menu ([supported agents](https://github.com/vercel-labs/skills#supported-agents)). It installs for all your projects.
  71. In a terminal, the command shows the agent menu. Without an interactive terminal, it installs into every agent folder, so use `-a <agent>` in scripts.
  72. <a href="https://skills.sh/aboudjem/humanizer-skill"><img src="https://img.shields.io/badge/skills.sh-000000?style=flat-square&logo=vercel&logoColor=white" alt="skills.sh" height="24"></a> Humanizer's page: [skills.sh/aboudjem/humanizer-skill](https://skills.sh/aboudjem/humanizer-skill).
  73. **Or install straight into one agent.** Each command installs for all your projects.
  74. <p align="center">
  75. <a href="#install-codex"><img src="https://img.shields.io/badge/Codex-000000?style=flat-square&logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyBmaWxsPSJ3aGl0ZSIgcm9sZT0iaW1nIiB2aWV3Qm94PSIwIDAgMjQgMjQiIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyI%2BPHRpdGxlPk9wZW5BSTwvdGl0bGU%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%3D" alt="Codex" height="24"></a>
  76. <a href="#install-claude-code"><img src="https://img.shields.io/badge/Claude_Code-D97757?style=flat-square&logo=claude&logoColor=white" alt="Claude Code" height="24"></a>
  77. <a href="#install-cursor"><img src="https://img.shields.io/badge/Cursor-3B3B3B?style=flat-square&logo=cursor&logoColor=white" alt="Cursor" height="24"></a>
  78. <a href="#install-gemini-cli"><img src="https://img.shields.io/badge/Gemini-1A73E8?style=flat-square&logo=googlegemini&logoColor=white" alt="Gemini CLI" height="24"></a>
  79. <a href="#install-github-copilot"><img src="https://img.shields.io/badge/Copilot-24292F?style=flat-square&logo=githubcopilot&logoColor=white" alt="GitHub Copilot" height="24"></a>
  80. <a href="#install-opencode"><img src="https://img.shields.io/badge/OpenCode-4A4A4A?style=flat-square&logo=opencode&logoColor=white" alt="OpenCode" height="24"></a>
  81. </p>
  82. <a id="install-codex"></a>**Codex**, tested:
  83. ```bash
  84. npx skills add Aboudjem/humanizer-skill -g -a codex
  85. ```
  86. <a id="install-claude-code"></a>**Claude Code**, install tested:
  87. ```bash
  88. npx skills add Aboudjem/humanizer-skill -g -a claude-code
  89. ```
  90. <a id="install-cursor"></a>**Cursor**, install tested:
  91. ```bash
  92. npx skills add Aboudjem/humanizer-skill -g -a cursor
  93. ```
  94. <a id="install-gemini-cli"></a>**Gemini CLI**, install tested:
  95. ```bash
  96. npx skills add Aboudjem/humanizer-skill -g -a gemini-cli
  97. ```
  98. <a id="install-github-copilot"></a>**GitHub Copilot**, install tested:
  99. ```bash
  100. npx skills add Aboudjem/humanizer-skill -g -a github-copilot
  101. ```
  102. <a id="install-opencode"></a>**OpenCode**, install tested:
  103. ```bash
  104. npx skills add Aboudjem/humanizer-skill -g -a opencode
  105. ```
  106. **Claude Code plugin.** The same skill, from the 10x marketplace. It adds no command or hook:
  107. ```bash
  108. claude plugin marketplace add Aboudjem/10x && claude plugin install humanizer@10x
  109. ```
  110. Only this project: drop `-g`. Another agent: find its `-a` code in [docs/editors.md](docs/editors.md).
  111. <details>
  112. <summary>Install without an installer (curl)</summary>
  113. This copies the one file the skill needs into your project:
  114. ```bash
  115. mkdir -p .claude/skills/humanizer && curl -sL https://raw.githubusercontent.com/Aboudjem/humanizer-skill/main/skills/humanizer/SKILL.md -o .claude/skills/humanizer/SKILL.md
  116. ```
  117. Swap `.claude/skills/` for the folder your agent reads. Use `~/.claude/skills/` for a global install.
  118. </details>
  119. ## Use it
  120. **1. Rewrite a draft.** In your agent, call the skill on your own file:
  121. ```text
  122. /humanizer --file draft.md --voice technical
  123. ```
  124. > **Before:** This comprehensive guide delves into the intricacies of our authentication system.
  125. >
  126. > **After:** This guide explains how our authentication system works.
  127. The rewrite cuts the AI habits. It never adds a fact the source did not have.
  128. By default it prints the rewrite in chat and leaves the file as it is. Add `--mode edit` to change the file in place.
  129. **2. Score a file.** The CLI is optional and runs from a clone. It reads the file and changes nothing:
  130. ```bash
  131. git clone https://github.com/Aboudjem/humanizer-skill && cd humanizer-skill
  132. node cli/index.js score examples/blog-post/before.md
  133. ```
  134. The output is the report under [What you get](#what-you-get). Use the path of your own file.
  135. **3. Check the rewrite kept the facts.** The real risk is a rewrite that drops a number:
  136. ```bash
  137. node cli/index.js compare --before examples/blog-post/before.md --after examples/blog-post/after.md --check-facts
  138. ```
  139. ```text
  140. Score: 57 -> 6 (delta -51)
  141. ... then 6 metric lines
  142. Fact check:
  143. OK. All 7 fact(s) in the original survive the rewrite.
  144. ```
  145. It exits 1 and names anything lost.
  146. **4. Stop bad drafts in CI** (your automatic checks). `scan --fail-above 40` exits 1 when a draft scores too high. The first output line is `Result: PASS` or `Result: FAIL`:
  147. ```bash
  148. node cli/index.js scan docs --fail-above 40
  149. ```
  150. ## How it works
  151. Your README sounds like a chatbot wrote it. You cannot say which words cause it.
  152. AI-written text has habits you can measure. Every sentence has about the same length. The same safe phrases appear again and again.
  153. | Part | What it does | Where it runs |
  154. | :-------- | :----------------------------------------------------------------------------- | :------------------- |
  155. | The skill | Names 55 patterns, finds them in your text, and rewrites in one of five voices | Your agent's model |
  156. | The CLI | Computes the 0 to 100 score from four signals | Node on your machine |
  157. The skill is one Markdown file, [`SKILL.md`](skills/humanizer/SKILL.md). Your agent reads it when you type `/humanizer`.
  158. The CLI score is plain arithmetic, so the same file always gets the same number:
  159. | Signal | Weight | What costs points |
  160. | :--------- | :----- | :---------------------------------------------------------- |
  161. | lexical | 0.40 | Stock AI words and phrases, such as "delve" or "in today's" |
  162. | burstiness | 0.28 | Sentences that are all about the same length |
  163. | diversity | 0.18 | A small vocabulary that keeps repeating |
  164. | repetition | 0.14 | The same three-word phrases coming back |
  165. _Burstiness_ is how much your sentence lengths vary. An _AI tell_ is one of those giveaway habits.
  166. <picture>
  167. <source media="(prefers-color-scheme: dark)" srcset=".github/assets/demo-burstiness-dark.svg">
  168. <source media="(prefers-color-scheme: light)" srcset=".github/assets/demo-burstiness-light.svg">
  169. <img alt="Sentence-length chart. The AI line is flat and uniform. The human line runs from 3 to 31 words. Humanizer restores the variation." src=".github/assets/demo-burstiness-light.svg" width="100%">
  170. </picture>
  171. <details>
  172. <summary>All 55 patterns, by category</summary>
  173. | IDs | Category | Examples |
  174. | :------ | :----------------- | :-------------------------------------------------------------------------------- |
  175. | P1-P8 | Content | Significance inflation, promotional language, AI vocabulary ("delve", "leverage") |
  176. | P9-P18 | Language and style | Negative parallelisms, dash overuse, structured-list syndrome |
  177. | P19-P21 | Communication | Chatbot artifacts, knowledge-cutoff disclaimers, sycophantic tone |
  178. | P22-P30 | Filler and hedging | Filler phrases, generic conclusions, uniform sentence length |
  179. | P31-P43 | Emerging | Elegant variation, placeholder text, chatbot markup leaks, infomercial hooks |
  180. | P44-P55 | Craft and forensic | False agency, aphorism formulas, unicode obfuscation, leftover hedge debris |
  181. Every pattern has its trigger list in [`SKILL.md`](skills/humanizer/SKILL.md). The deep dives are in [`references/patterns.md`](skills/humanizer/references/patterns.md).
  182. The core catalog (P1-P30) draws on [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) (CC BY-SA).
  183. </details>
  184. ## Works with
  185. | Agent | Install | Check that ran | Tested |
  186. | :------------------ | :------------------------------------------------------ | :-------------------------------------------------------------- | :--------- |
  187. | Claude Code 2.1.326 | `claude plugin install humanizer@10x` | Installed for the user from the 10x marketplace, and validated in strict mode | 2026-11-17 |
  188. | Codex CLI 0.169.0 | `npx skills add Aboudjem/humanizer-skill -g -a codex` | Installed, and the skill rewrote a sentence (project install) | 2026-11-17 |
  189. | Gemini CLI | `npx skills add Aboudjem/humanizer-skill -g -a gemini-cli` | Install only. The skill file landed in `~/.agents/skills/` | 2026-11-17 |
  190. | Cursor | `npx skills add Aboudjem/humanizer-skill -g -a cursor` | Install only. The skill file landed in `~/.agents/skills/` | 2026-11-17 |
  191. | GitHub Copilot | `npx skills add Aboudjem/humanizer-skill -g -a github-copilot` | Install only. The skill file landed in `~/.agents/skills/` | 2026-11-17 |
  192. | OpenCode | `npx skills add Aboudjem/humanizer-skill -g -a opencode` | Install only. The skill file landed in `~/.agents/skills/` | 2026-11-17 |
  193. Each global install was run into an empty home folder on 2026-11-17. Documented, not tested: Windsurf, Zed, Kimi Code CLI and the other agents in [docs/editors.md](docs/editors.md).
  194. The skill is Markdown, so it runs on whatever model your agent uses. It is not tied to one model.
  195. ## Proof
  196. Every number on this page comes from a command you can run in a clone.
  197. | Claim | Command | It prints |
  198. | :---------------------------------------------- | :--------------------------------------------------------------------------------- | :----------------------------------- |
  199. | 90 tests pass | `npm test` | `# pass 90` and `# fail 0` |
  200. | 55 patterns | `grep -cE '^\*\*P[0-9]+: ' skills/humanizer/SKILL.md` | `55` |
  201. | The repo passes its own gate | `npm run selfcheck` | `Result: PASS` three times |
  202. | The example goes from 57 to 6 and keeps 7 facts | the `compare` command in [Use it](#use-it) | `Score: 57 -> 6` and `All 7 fact(s)` |
  203. | The CLI has no dependency | `node -p "Object.keys(require('./cli/package.json').dependencies \|\| {}).length"` | `0` |
  204. - CI runs the tests on Node 18 and 20, on every push and pull request.
  205. - CI also runs both install paths: `npx skills add . --list` and `claude plugin validate . --strict`.
  206. - `npm run selfcheck` scores this repo with its own CLI. It also rejects an em dash in any doc.
  207. ## FAQ
  208. ### Does anything leave my machine?
  209. Not from Humanizer itself. The skill and the CLI make no network call of their own. There is no telemetry, no account and no API key.
  210. The rewrite runs on your agent's model. If that model is hosted, your text goes to it, as with any prompt. The CLI score runs on your machine only.
  211. ### Will it make my text pass an AI detector?
  212. That is the wrong goal, and no honest tool can promise it. Detectors change often, and nobody agrees on how well they work on rewritten text. Writing with real variation and concrete detail reads better to people.
  213. ### It flagged text I wrote myself. Is it broken?
  214. Probably not, but it can be wrong about you. Some people write even sentences by habit. Detectors also make more mistakes on non-native English writing ([Liang et al.](https://arxiv.org/abs/2304.02819)). Treat the score as a signal.
  215. ### Which models does it run on?
  216. Whichever one your agent uses. The runs on this page used Claude Sonnet in Claude Code and the default model in Codex CLI.
  217. ### Do I need the CLI?
  218. No. The skill alone rewrites text and estimates a score. The CLI exists for a number that never changes between runs. That is what a CI gate needs.
  219. ### Is the CLI on npm?
  220. No. Run it from a clone with `node cli/index.js`. It needs Node 18 or newer and no install step.
  221. The `npx skills` installer is a separate tool. It needs Node 22 or newer.
  222. More answers are in [docs/faq.md](docs/faq.md).
  223. ## Limits
  224. - **It is not a detector.** The score reads four text signals. It cannot prove who wrote a text.
  225. - **It does not promise to pass a detector.** It edits the writing, nothing more.
  226. - **The CLI metrics are built for English.** The Chinese pattern appendix is a provisional draft.
  227. - **The skill's own `--score` is a model estimate.** Only the CLI number is the same on every run.
  228. - **The fact check compares tokens, not meaning.** It misses two swapped numbers and a changed claim with no number in it.
  229. - **A sample under 40 words is low-confidence.** `scan` lists it and never fails on it.
  230. - **The rewrite depends on your model.** A weaker model follows the rules less closely.
  231. ## Learn more
  232. - [The skill itself](skills/humanizer/SKILL.md), and the [pattern deep dives](skills/humanizer/references/patterns.md)
  233. - [What the score measures](docs/science.md), the research behind the rewrite rules
  234. - [Install in your editor](docs/editors.md), [gate commits](docs/pre-commit.md), [how it compares](docs/comparison.md)
  235. - [The metrics CLI](cli/README.md), the [CHANGELOG](CHANGELOG.md), the [LICENSE](LICENSE)
  236. <p align="center">
  237. <sub>Built by <a href="https://github.com/Aboudjem">Adam Boudjemaa</a> · MIT License · No telemetry, no data collection</sub>
  238. </p>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.