Libraries and authors¶
The examples use arp as shorthand for python <skill-directory>/scripts/arp.py.
Local PDF/BibTeX libraries¶
arp library audit <library-directory> --output <audit-directory>
arp library update <library-directory> --output <staging-directory>
arp library import <incoming-directory> --destination <library-directory> --output <plan.json>
arp library organize <library-directory> --output <plan.json>
arp library apply --plan <plan.json>
arp library apply --plan <plan.json> --rollback
Library audits report identities, duplicates, missing/ambiguous pairs, citation placement, and the PDF's independently evidenced version. A current BibTeX record does not prove the PDF is current. Registry metadata stays discovery evidence.
Update stages official pairs; import/organize produce hash-bound plans; explicit apply checks hashes and keeps backups. Raw BibTeX bytes, BOM, encoding, and line endings are preserved. Optional --metadata inputs enable offline comparison; update --verify-only avoids acquisition. A bounded --watch observes user-downloaded ScienceDirect pairs and maintains one queue without automating a browser. See the library reference.
Author bibliographies¶
Author specs accept identity IDs, affiliation/coauthors, date range, homepage, and a local library. Public ORCID, DBLP, homepage, and database records retain identity uncertainty and source errors. Exact DOI deduplication, formal/early-access/preprint groups, issue dates primary to online dates, and explicit version relations remain auditable; coverage is never claimed absolutely complete. --records supplies offline source evidence. See the author reference.