SkillWorks: Scored Claude Code skills, subagents and plugins

Maker here. Every list of Claude Code skills I could find was a hand-written README, which means it is a snapshot of what one person happened to notice, and it is wrong the week after it is published. SkillWorks reads the repositories instead and rebuilds the whole index nightly. Each listing gets one 0-100 number from four measured components — Works 40%, Maintained 25%, Adopted 20%, Documented 15% — and the page shows you the four numbers beside the total, so you can disagree with the weighting and still use the parts. Works is the one that matters: does the frontmatter parse, do the allow-listed tools actually exist in Claude Code, do the referenced files exist in the repo, is the body more than a stub. Tens of thousands of listings fail that, including some with very large install counts, and there is a whole page for them with the exact check each one failed. Nothing is hand-picked and there is no way to pay for a position. Happy to answer anything about the scoring.