A per-link database of every anchored hyperlink across all 34 substantive repositories in the leesharks000 network — 59,431 rows, each carrying anchor text, current href, surrounding prose context, extracted DOI/AXN references, resolver lookup, prose-vs-resolver-title jaccard, best-matching alternative registry record, and a verdict flag classifying each link into review categories (no_id, ambiguous, consistent, href_disagrees_resolver, external_ok, href_wrong_target, likely_prose_mismatch, ok). Companion to DOI Resolution Index (#4) — this inventory audits the resolver's real-world usage across the site network. Assembled to answer a question the earlier 440K-row link-token scan could not: not just how many links there are, but whether each link is right. Immediate use is a review pile for correcting broken and mismatched links across sites, with corrections applied preferentially to the resolver map at its source of truth (deposit #4 v3.11.0+) rather than to individual site source files, so sovereign resolver behavior propagates automatically via /go/?doi=X. Artifacts at /data/link-inventory/: inventory-summary.json (aggregate stats), link-review-actionable.csv (623 rows, high-confidence mismatches), link-review-ambiguous.csv (830 rows, DOI-carrying ambiguous cases with promising alt). Regeneration scripts at /scripts/link-inventory/. Iteration policy: rescan after each rewrite pass or resolver update; bump minor version; the versioned snapshots become the durable record of network-wide link integrity. v1.1 (2026-07-08): 106 fixes applied since v1.0 — 69 sentinel-0 hrefs corrected to specific records via unique-substring anchor-text matching, 34 href_disagrees_resolver corrections, 3 ambiguous fixes. Verdict deltas: href_wrong_target 258→208, ambiguous 11163→11149, consistent 1286→1303. v1.2 (2026-07-08): additional fix rounds. Total inventory-driven fixes this session: 448 (69 sentinel-0 unique-substring + 34 disagrees verified + 3 ambiguous + 127 loose sentinel-0 + 126 loose disagrees + 40 ambiguous loose + 49 relaxed-criteria final). Plus 8,462 earlier Phase 10 rewrites. Final verdict deltas from v1.0: href_wrong_target 258→87 (-171), ambiguous 11163→11124 (-39), consistent 1286→1322 (+36), href_disagrees_resolver 365→569 (+204 — mostly analyzer sensitivity from fixes cascading into DOI-based mismatch flags that were not previously visible because href was sentinel). Known source-side editorial issue documented: lagrange-observatory Generative Monoculture DOI 20675438 typo (should be 20675199 or 20674488) — resolver mapping is correct, source HTML has wrong DOI hardcoded. v1.3 (2026-07-08): 827 per-citation DOI-typo corrections. Cross-clustering by (wrong_doi, correct_doi) revealed 478 unique off-by-N pairs, most differing by ±1 in the DOI number (systematic citation typos from adjacent Zenodo mints). Fixes applied per-file:line — each rewrite touches exactly the one citation where prose describes work Y but anchor cites a DOI resolving to work Z. Preserves legitimate citations of the same DOI at pages where it correctly cites its true target. V6 global DOI-swap pass was rolled back after breaking such legitimate citations. v1.4 (2026-07-08): Phase 14 comprehensive sweep. 750 additional fixes: 88 multi-signal anchor+heading+prose matches, 532 context-DOI /go/ routing (for anchor links where the DOI appears in adjacent prose/table cell rather than in anchor text), 130 residual mixed. Final actionable state ~5% of v1.0 count.
A per-link database of every anchored hyperlink across all 34 substantive repositories in the leesharks000 network — 59,431 rows, each carrying anchor text, current href, surrounding prose context, extracted DOI/AXN references, resolver lookup, prose-vs-resolver-title jaccard, best-matching alternative registry record, and a verdict flag classifying each link into review categories (no_id, ambiguous, consistent, href_disagrees_resolver, external_ok, href_wrong_target, likely_prose_mismatch, ok). Companion to DOI Resolution Index (#4) — this inventory audits the resolver's real-world usage across the site network. Assembled to answer a question the earlier 440K-row link-token scan could not: not just how many links there are, but whether each link is right. Immediate use is a review pile for correcting broken and mismatched links across sites, with corrections applied preferentially to the resolver map at its source of truth (deposit #4 v3.11.0+) rather than to individual site source files, so sovereign resolver behavior propagates automatically via /go/?doi=X. Artifacts at /data/link-inventory/: inventory-summary.json (aggregate stats), link-review-actionable.csv (623 rows, high-confidence mismatches), link-review-ambiguous.csv (830 rows, DOI-carrying ambiguous cases with promising alt). Regeneration scripts at /scripts/link-inventory/. Iteration policy: rescan after each rewrite pass or resolver update; bump minor version; the versioned snapshots become the durable record of network-wide link integrity. v1.1 (2026-07-08): 106 fixes applied since v1.0 — 69 sentinel-0 hrefs corrected to specific records via unique-substring anchor-text matching, 34 href_disagrees_resolver corrections, 3 ambiguous fixes. Verdict deltas: href_wrong_target 258→208, ambiguous 11163→11149, consistent 1286→1303. v1.2 (2026-07-08): additional fix rounds. Total inventory-driven fixes this session: 448 (69 sentinel-0 unique-substring + 34 disagrees verified + 3 ambiguous + 127 loose sentinel-0 + 126 loose disagrees + 40 ambiguous loose + 49 relaxed-criteria final). Plus 8,462 earlier Phase 10 rewrites. Final verdict deltas from v1.0: href_wrong_target 258→87 (-171), ambiguous 11163→11124 (-39), consistent 1286→1322 (+36), href_disagrees_resolver 365→569 (+204 — mostly analyzer sensitivity from fixes cascading into DOI-based mismatch flags that were not previously visible because href was sentinel). Known source-side editorial issue documented: lagrange-observatory Generative Monoculture DOI 20675438 typo (should be 20675199 or 20674488) — resolver mapping is correct, source HTML has wrong DOI hardcoded. v1.3 (2026-07-08): 827 per-citation DOI-typo corrections. Cross-clustering by (wrong_doi, correct_doi) revealed 478 unique off-by-N pairs, most differing by ±1 in the DOI number (systematic citation typos from adjacent Zenodo mints). Fixes applied per-file:line — each rewrite touches exactly the one citation where prose describes work Y but anchor cites a DOI resolving to work Z. Preserves legitimate citations of the same DOI at pages where it correctly cites its true target. V6 global DOI-swap pass was rolled back after breaking such legitimate citations. v1.4 (2026-07-08): Phase 14 comprehensive sweep. 750 additional fixes: 88 multi-signal anchor+heading+prose matches, 532 context-DOI /go/ routing (for anchor links where the DOI appears in adjacent prose/table cell rather than in anchor text), 130 residual mixed. Final actionable state ~5% of v1.0 count.