Human Writing

Problem

Model-written articles often sound “correct but cold”: abstract claims, tidy logic, and no embodied experience. They also tend to rely on 1. 2. 3. or “First, second, third” style lists. This skill treats that bullet-point habit as one of the main signals of an AI-like tone. Its goal is not to make text more document-like, but to make it read as if a person with concrete experience, preferences, and stakes wrote it.

How It Works and Where to Be Careful

It supports writing mode and polishing mode. Writing mode starts from a topic or request, drafts the article, checks for AI-like patterns, confirms the structure is paragraph-based, and then refines it. Polishing mode analyzes existing text, prioritizes converting bullets, numbered lists, and summary phrases into flowing paragraphs, and then applies style adjustments.

The core methods are:
- Embodiment: add specific experience, bodily feeling, and scene detail.
- Imperfection: keep uncertainty such as “I think”, “maybe”, or “not for everyone”.
- De-definition: replace abstract definitions with scenarios and stories.
- Consequence: name risks, costs, and reasons for a choice.
- Personal adaptation: store preferences such as formality, common words, and emotional tone in ~/.human-writing-config.json.

It fits blog posts, copy, emails, or other prose that needs a natural voice, and it also helps when existing text feels templated. Be careful not to use it to replace factual accuracy in technical or professional writing; if the user needs a strict formal structure, preserve clarity and do not force slang or over-personalization.