What problem does it solve? Deterministic checks can catch typos and stale bios, but they cannot judge whether a public README or positioning doc actually reads professionally. This Skill fills that gap by scoring prose quality and recording a verdict pinned to the exact bytes reviewed, so edits automatically trigger re-review. ## Core Features & Use Cases - Four-Axis Voice Scoring: Grades prose 0-5 on tone, brand voice, clarity, and credibility from the perspective of investors, hiring managers, and customers. - Content-Pinned Verdicts: Records each judgment with the SHA-256 hash of the exact bytes reviewed, so any later edit invalidates the verdict and re-queues review. - Keeper Integration: Works with decorum-keeper.py, which detects changed prose surfaces, queues voice-judge findings, and clears them once a matching verdict is appended to decorum-judgments.yaml. - Use Case: After rewriting a repository README, run the keeper sweep, score the new prose against the four axes, and append a pass/fail verdict with a specific suggested fix for any defects. ## Quick Start Ask the agent to run the decorum keeper sweep, review the queued prose surface, and record a voice verdict pinned to its current content hash.
