evaluation-anchor-checker

Audit and rewrite evaluation/numeric claims to ensure they carry minimal protocol context (task + metric + constraint) and avoid underspecified model naming.

Independent GitHub snapshot · Automated classification · Original documentation governs