What problem does it solve? AI-generated and hand-drafted copy carries recognizable writing tells — antithesis reflexes, rule-of-threes lists, filler words, US spelling — that survive a casual read and erode a brand's voice. This Skill audits an existing draft against a structural AI-pattern checklist and the Lia brand-voice standard, then reports every issue before the copy ships. ## Core Features & Use Cases - Structural AI-pattern scan: Runs 32 documented checks (negative parallelism, compulsive summaries, em-dash overuse, significance inflation, and more) across the full draft, logging every match with location and suggested fix. - Deterministic word-check backstop: A Python script flags hard-avoid vocabulary, US spellings, sense-dependent words (license, meter, dialog), watchlist terms, and filler, with a self-test mode and JSON output. - Brand and product-voice quality checks: Qualitative pass/fail reads against the Lia base voice and product profiles like Held, covering vocabulary rules and tone. - Use Case: Before shipping onboarding copy for a Lia product, run the audit on the draft to get a structured report of AI tells, voice drift, and spelling issues — then optionally apply surgical fixes in a hard pass. ## Quick Start Ask the agent to run a voice check on your drafted copy, for example by saying "voice-check this draft of our release notes against the Lia brand voice."
