* fix(learning): preserve candidate identity and research feedback

* docs(learning): remove obsolete architecture verdicts

* docs(learning): separate learning claims and evidence

* docs(learning): refine audit source anchors

* chore(release): align the repaired learning packages

* docs(learning): retain audit findings and release evidence