humanize-text-prolisted
Install: claude install-skill sarrafgsarraf/humanize-text-pro
# Humanize Text Pro
One editor, one ruleset. This skill merges three upstream skills that each solved
part of the problem and disagreed on the rest:
| Source | What it contributes |
|---|---|
| [petergyang/no-ai-slop](https://github.com/petergyang/no-ai-slop) | Voice preservation, minimum effective edit, slop patterns, a pass/fail self-check |
| [blader/humanizer](https://github.com/blader/humanizer) | The Wikipedia "Signs of AI writing" catalog, false-positive discipline, human details worth keeping |
| [danyuchn/asd-ste100-skill](https://github.com/danyuchn/asd-ste100-skill) | ASD-STE100 structural clarity: one meaning per word, one instruction per sentence, modality preservation |
They conflict in fifteen places. Every conflict is settled once, in
[Conflict rulings](#conflict-rulings). **Those rulings supersede the upstream
wording in all three source skills.** Do not go back to a source rule that a
ruling has overridden, and do not apply two sources' contradictory advice in the
same pass.
## The four laws
These hold in every register, every mode, every time. Nothing below overrides them.
1. **Invent nothing.** Never add a fact, name, number, date, statistic, quote,
citation, source, cause, frequency, or ranking that is not in the source or
supplied by the user. If a sentence reads better because you supplied a
mechanism, you have stopped editing and started writing fiction. Ask for the
missing detail or use a simpler sentence. (Fiction and creative writi