What problem does it solve? Drafts written or edited by AI often ship with detectable machine-writing tics, inflated claims, and a voice that does not match the human author whose name goes on them. This Skill provides a measured, repeatable review protocol that catches those problems before content ships. ## Core Features & Use Cases - Measured AI-tell sweep: Run scripts/tic-count.py to count em-dashes, antithesis patterns, ceremony phrases, and other tells per 100 words, with concrete thresholds from a real 8,742-word draft. - Claims discipline and voice matching: Verify every number and capability claim against evidence, and calibrate rewrites against verbatim samples of the real author's diction. - Parallel multi-lane review protocol: Freeze a pinned draft, fan out to independent reviewers across technical, structural, security, and voice lanes, then reconcile findings with a tracking note. - Use Case: Before publishing a proposal or blog post drafted with AI assistance, run the tic-count script, audit every statistic against its source, and produce before/after rewrites of the five worst passages in the owner's voice. ## Quick Start Review the attached draft of my announcement post for AI-writing tells, unsupported claims, and voice drift, then give me the five worst passages rewritten in my voice.
