Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Saturday, technology Businessman Siqi Chen Released An open source plugin for Anthropic Claude Code An AI assistant that instructs the AI model to stop typing like the AI model.
The simple plugin called Humanizer provides Claude with a list of 24 language and formatting patterns used by Wikipedia editors. included As a gift for chatbots. Chen posted the plugin on GitHub, where it had captured more than 1,600 stars as of Monday.
“It’s really helpful that Wikipedia compiled a detailed list of ‘AI writing tags,'” Chen said books On X. “So much so that you can ask your LLM not to do it.”
The source material is a guide from Clean Artificial Intelligence WikiProjecta group of Wikipedia editors who have been searching for AI-generated articles since late 2023. French Wikipedia editor Elias Lebleau founded the project. Volunteers flagged more than 500 articles for review, and in August 2025, published An official list of the patterns they kept seeing.
qin he toolSkill file” for Claude Code, Anthropic’s terminal coding assistant, which includes a Markdown file that adds a list of typed instructions (you can See them here) is appended to the prompt that is fed into the large language model that runs the helper. Unlike normal System administratorFor example, skill information is formatted in a standardized way so that CLOUD models are fine-tuned to interpret it more accurately than a regular system prompt. (Custom skills require a paid Claude subscription with code execution turned on.)
But as with all AI claims, language models don’t always track skill profiles perfectly, so does Humanizer actually work? In our limited tests, Chen’s skill profile made the AI agent’s output look less accurate and more casual, but it may have some drawbacks: it won’t improve realism and may hurt programming ability.
In particular, some Humanizer instructions may lead you astray, depending on the task. For example, the Humanizing skill includes this line: “You have opinions. Don’t just report facts, engage with them. ‘I really don’t know how to feel about this’ is more humanizing than neutrally listing pros and cons.” Although imperfection seems human, this kind of advice probably won’t help you if you’re using Claude to write technical documentation.
Even with its flaws, it’s ironic that one of the web’s most referenced rulesets for AI-assisted typing detection might help some people subvert it.
Discover patterns
So what does AI writing look like? Wikipedia’s guide is specific and contains many examples, but we’ll give you just one here for the sake of brevity.
Some chatbots like to pepper their topics with phrases like “identify a pivotal moment” or “stands a testament to that,” according to the guide. They write like tourist brochures, describing the views as “breathtaking” and describing towns as “lies within” scenic areas. They add “-ing” phrases to the end of sentences to make them sound analytical: “Symbolizes the region’s commitment to innovation.”
To get around these rules, the Humanization skill asks Claude to replace inflated language with plain facts and offers this example of a conversion:
before: “The Statistical Institute of Catalonia was officially created in 1989, representing a pivotal moment in the development of regional statistics in Spain.”
after: “The Statistical Institute of Catalonia was created in 1989 to collect and publish regional statistics.”
Claude will read that and do its best as a pattern-matching machine to generate output that matches the context of the conversation or task at hand.
Why AI typing detection fails
Even with this confident set of rules formulated by Wikipedia editors, we did it Previously written On why AI writing detectors don’t work reliably: There is nothing inherently unique about human writing that reliably distinguishes it from LLM writing.
One reason is that although most AI language models tend toward certain types of language, they can also be asked to avoid them, as is the case with the humanization skill. (Although sometimes it’s very difficult, as OpenAI found on its site A struggle that lasted years v. dash m.)