use-smart-humanize-text — eval reports

Two scenarios drive the published evidence for this skill. Each scenario page below surfaces the same fields in the same order — metadata strip, what we're testing, why, headline numbers, comparison deltas, link to the full interactive report, and an inline live preview iframe.

Scenario	Skill arms	Headline pass rate	Runner
Adaptive brand voice discovery	smart / flat / none	smart 0.935 vs flat 0.679	cursor-agent
Indie-launch copy iteration	smart / flat_wiki / flat / none	smart 0.873 vs flat_wiki 0.842 (S5/S6 retention: smart 0.922 / 0.944)	cursor-agent

Each .html artifact embedded on those pages is a self-contained Plotly dashboard generated by humblskills eval run. Every published report uses a dated filename — there is no latest-* alias.