jev-writerlisted

# Analysing writing against real engagement and craft rubrics This skill turns "why do some of my posts work?" and "is my technical writing rigorous and free of slop?" into answers grounded in real data and calibrated rubric evaluations. You can evaluate a whole corpus of published writing or run a fast pre-flight audit on an unpublished draft (`.md` or `.txt`). **The whole value is honesty.** A dozen tools already tell people their content is great. This one tells them when their sample is too small to know, when a rubric is broken, when a technical post lacks checkable proof artifacts, and when the thing they are proud of predicted nothing. Do not soften that. A flattering answer is a useless answer. ## What this cannot do Say this early and plainly if the user seems to expect otherwise: - **It does not write or rewrite posts.** Jev is a System One model and is not trained to generate text. It returns typed judgments and probabilities. - **It cannot recover engagement data the platform no longer exposes.** On LinkedIn, impressions exist for roughly the last 12 months and the top 50 posts. Older engagement is gone for everyone, including the platform's own UI. - **It cannot establish causation.** Everything it finds is an association in one person's history, with unmeasured confounders. ## Step 1. Select the content type and rubric pack `jev-writer` ships with four calibrated rubric packs. Always identify which format the user is evaluating: | Format / Goal |