[Darssal]

MemLab and the recruitment classifier now have verified standalone repositories. This change makes Research-Code-Collection a documentation index: current project source moves out of the default branch while its existing Git history remains available for traceability.

The README points to FracLogGPT-Lite, MemLab, Recruitment-Text-Classification and JSP-UI-Prototype as the canonical maintenance destinations. Engineering notes retain their design explanations and link to independent source/tests. Source attribution, publication scope and contribution rules now follow that structure. The former classification/memory CI jobs are replaced with an index inventory, canonical link validation, publication scans and maintenance regressions.

Validation:

• MemLab main 10c7cfe69fdb76c0b7dcae5b83398e25892302e5: existing 53 offline tests and Linux CI verified.
• Classification main b0226a9a69e24482b52a4931fb12ffe823c5bb25: 16 classification regressions, 16 Linux publication regressions and synthetic CPU demo passed; both CI jobs succeeded.
• JSP main 8b74139a1bec353d25666b6642d0718058cc5770: all 13 JSP and original documentation Git blobs preserved, 18 navigation targets checked, three legacy-encoding boundary regressions and Linux CI passed. CodePen attribution and prototype/no-backend scope retained.
• This branch: 114 duplicated project files removed; exact 12-file documentation/maintenance inventory checked; four public repositories and 36 canonical destinations verified using read-only GitHub API calls.
• Working-tree and actual Git-index publication scans: zero findings. Local maintenance suite: 16 passed, one Windows symlink skip; Linux CI exercises that test.

The public baselines and source checks do not reproduce paper metrics or verify live services, databases, BERT training, JSP deployment or authentication. No raw data, private research history, model files or personal evidence is imported. No repository is deleted or changes visibility, and no permissions are expanded.