responseAware

standard plus AnswerAnchorGuard, the one guard that reads the candidate's stored answer.

Opt in deliberately. Every guard in standard is measured on corpora written before it existed; this one is measured on answers written for the purpose, because no corpus of real paired answers exists to measure it against. The number is a regression check rather than a blind measurement, and mixing the two under one default would quietly downgrade the evidence behind everything else. docs/CORPUS.md states the difference and ResponseGuardTest holds the numbers.

It costs nothing extra to run: the response is already on the candidate entry, and the guard only reads it once the prompts have proved to differ by a substitution.

responseAware, reading every marker from vocabulary.