Humanizer: Remove AI-Generated Writing

Problem

AI-generated prose often uses templated connectors, inflated symbolism, triadic praise, and vague conclusions, such as serves as a testament, Moreover, and seamless, intuitive, and powerful. These patterns make technical notes, product feedback, or engineering documentation sound promotional. @zcwl/humanizer is not a fact rewriter; it is a style-focused text processor intended to reduce obvious AI traces and bring output closer to human-written text.

How it works

The skill centers on text processing and accepts parameters such as content, mode, and style. content is the input text; mode selects the output shape, with values including json, text, and markdown; style controls the output style. During processing, it typically removes empty attribution, templated syntax, over-ornamentation, and superficial analysis, then replaces abstract claims with concrete facts, feature points, or verifiable information. For example, the provided case replaces a promotional sentence with adds batch processing, keyboard shortcuts, and offline mode.

Boundaries

The documentation indicates that it currently focuses mainly on Chinese text and is not suitable for complex decisions that require human judgment. Inputs should not contain sensitive information. Batch task orchestration, advanced rules, export and multi-format conversion, and monitoring alerts are listed as paid extended capabilities. It is useful for removing templated tone from technical notes and feedback. Check the execution status and re-verify key facts; do not treat it as automatic fact-checking.