What problem does it solve? AI-generated prose carries predictable tells: filler phrases, formulaic structures, vague declaratives, and marketing puffery that make technical writing read as machine output. This Skill detects and rewrites those patterns in documentation, ADRs, PR bodies, and scientific writing while preserving technical accuracy and domain terminology. ## Core Features & Use Cases - Pattern catalogs: Five reference files catalog phrases, structures, tropes, and before/after examples covering throat-clearing openers, binary contrasts, dramatic fragmentation, false agency, and tech puffery. - Scoring rubric: Rates text 1-10 across directness, rhythm, trust, authenticity, and density; below 35/50 triggers revision. - Naming audit: Enforces domain-first naming for durable operator surfaces (recipes, paths, CLI commands) and distinguishes governance taxonomy citations from identity usage. - Git safety guards: Refuses to touch commit messages, run git commands, or mutate history unless the user explicitly opts in to desloping a draft commit message. - Use Case: Paste a draft ADR full of phrases like "seamlessly integrates" and "robust solution" and receive a rewritten version naming concrete databases, algorithms, and latency metrics, plus a naming audit table. ## Quick Start Ask the assistant to review this draft using the deslop checklist and rewrite it to pass the quick checks.
