What problem does it solve? Surface-level editing (fixing words, punctuation, and phrasing) leaves the structural fingerprint of AI writing intact: stated lessons, tidy single-track arcs, embodied-emotion clichés, vague allusions, and uniform shapes across pieces. Research cited in this Skill shows narrative structure alone detects AI text at 93.2% F1, while professional stylistic rewriting moved detection only 1.6 points. ## Core Features & Use Cases - Six structural audits: theme explicitness, structural tidiness, emotion mode, reference specificity, reader engagement, and shape convergence, each run one at a time against an extracted outline of the piece. - Deterministic scanner: scripts/structural_scan.py flags grep-able tells (embodied emotion, stated-lesson markers, tidy closers, vague allusions) with JSON output and a --strict mode for hooks. - Genre calibration and intervention menu: references/genre-calibration.md maps which audits apply to LinkedIn posts, course lessons, blog posts, and emails, and a rotating intervention menu prevents replacing one AI default with another. - Use Case: After running a surface humanizer pass on a LinkedIn post draft, extract its skeleton, run the six audits, pick one intervention (e.g., cut the restated takeaway and end on the number), then verify with python3 scripts/structural_scan.py draft.md. ## Quick Start Ask the AI to run the structural-humanizer pass on your draft, for example: "Run the structural pass on this blog post and remove the AI structural tells."
