A Claude Code humanizer is a writing skill you can run inside your coding workflow to clean up AI-shaped prose before it ships. This page explains what the popular blader/humanizer skill is, when the web tool is enough, and when to install the skill.
blader/humanizer is an open-source skill for contextual review of 35 writing patterns in Claude Code and compatible skill clients.
The upstream project is by Siqi Chen, GitHub user blader. Blader Humanizer is an unofficial web guide and tool inspired by that repo, not the official project.
The skill follows the same basic idea as the Wikipedia guide to signs of AI writing: look for generated-text habits, then rewrite toward plain, specific prose.
The web version and the Claude Code skill solve the same writing problem, but they fit different workflows.
| Use case | Web tool | Claude Code skill |
|---|---|---|
| Best for | fast check One-off drafts, emails, short pages, ChatGPT output, Gemini output, and quick copy edits. |
workflow fit Docs, README files, changelogs, launch copy, support macros, and repeated drafts inside Claude Code. |
| Setup | No install. Paste text, detect 12 concrete signals, review suggested edits, and copy the result. | Install the skill once, then call it from your editor or ask Claude Code to humanize a file. |
| Privacy | Runs in your browser and does not upload pasted text. | Runs in your Claude Code / OpenCode workflow. Your normal editor and model-session privacy rules apply. |
| Voice matching | Good for generic cleanup: fewer tells, clearer prose, less filler. | Better for personal voice calibration because you can provide style samples and run it across many drafts. |
The current official Skills CLI command installs the complete upstream skill globally:
For the current 35-pattern skill, follow the official installation guide or download the upstream source ZIP. This site no longer generates a smaller substitute skill.
/install is for installation intent: use the official Skills CLI or Claude Code plugin commands, or download the official source ZIP.
This page is for guide intent: understand what the Claude Code humanizer skill is, how it differs from the web tool, and how voice calibration changes the result.
The biggest reason to install the skill is not the first cleanup pass. It is calibration. Give the skill a short sample of your own writing, then ask it to rewrite the AI draft with that rhythm and level of bluntness.
For a developer, this matters on docs and launch notes: a clean but generic rewrite can still sound like model output. A calibrated rewrite can keep your shorter sentences, naming habits, technical caveats, and preferred level of detail.
A humanizer is useful editing help. It is not evidence laundering, and it should not replace judgment.
This site is not built to bypass Turnitin, GPTZero, or similar tools. Use it to make an assisted draft clearer before a person reads it.
Em dashes, triads, and polished transitions are not proof of AI authorship. Treat them as editing cues and judge each passage in context.
The last pass still belongs to a human. Check facts, source claims, delete vague phrasing, and keep the lines that actually sound like you.
Paste a draft for a quick 12-signal web check, or install the official 35-pattern skill from upstream for a full contextual review.
Unofficial and not a detector-bypass tool. This page is not affiliated with Siqi Chen, GitHub user blader, the upstream blader/humanizer repo, Anthropic, OpenAI, or Google. The checker is for clearer editing, not for misrepresenting authorship or evading AI-detection systems.
For the current Skills CLI and Claude Code plugin commands, plus links to the official repository and source ZIP, use the install page. For pattern definitions, use the AI Writing Patterns hub.
If you want to check a draft for literal writing signals before deciding on an edit, use the Signs of AI Writing checker. To return to the main tool, open Blader Humanizer.