text-humanizerlisted

# Text Humanizer Skill This skill helps users create content that sounds human, not algorithmic. It has three modes: 1. **Humanize**: Rewrites existing text to remove AI patterns using traffic-light diagnosis, contrastive subtraction, and a staged verification pass. 2. **Audit**: Diagnoses AI markers in text WITHOUT rewriting. Returns a traffic-light map and pattern list. 3. **Generate Prompt**: Creates a specialized System Prompt for a specific genre/domain that the user can use in other chats. ## Scope boundary — what this is NOT **Long-form fiction is out of scope.** A short story, a novel chapter, a screenplay: this skill edits the surface of a text — vocabulary, syntax, punctuation, paragraph rhythm — and in fiction that surface carries a small share of what marks a text as machine-written. Measured: a classifier using only discourse-level narrative features (plot continuity, thematic explicitness, how emotion is conveyed, chronology) separates human from AI fiction at 93.2% macro-F1, and stylistic span-rewriting of the kind this skill performs moved it by 1.6 points, from 95.5% to 93.9% (StoryScope, [arXiv:2604.03136](https://arxiv.org/abs/2604.03136); LAMP, CHI 2025). Running this skill on a story therefore returns a cleaner surface over the same narrative architecture. That is not nothing, and it is not what the request usually means. Say so, and offer the nearest thing that is in scope: `--genre blog` covers essayistic and personal writing, which is what `patt