Skip to content

align

ClaudePlugin

Personal evals for Claude Code and Cowork. Three skills: /align rates LLM claims (correct/wrong/almost/needs-nuance/can't-verify/skipped) in a local HTML form; /retro synthesizes the archive into failure-mode clusters + gated patches; /diagnose traces wrong claims back to the stale instruction that caused them.

Best rank—Not in a ranked chart
Charts—
LaunchedMay 26
Not in a Claude chart we track