[docxology]

Coordinated release verification — not publication acceptance

• Candidate AGEINT source: a3cea04fbf2dd73c419fe9fe3506ea57ea9b7bd9. Official CFF1.2 validation failed before the correction and passes afterward; release version0.2.0, existing concept DOI and both licenses are unchanged.
• Original H6 whole-book run:475 tests pass at90.31% coverage;333 Markdown files,3440 supported headings and1826 PDF pages. All9 SDK gates and rendered-reference checks pass. Aggregate readiness correctly refuses unfinished task36/37 prerequisites.
• Original H6/B2 inspection reader was exercised at390px and1280px, including painted deep References navigation, offline back-navigation, all177 images and13 MathJax containers with0 reported math errors. This is original inspection evidence, not post-readiness promotion or final-source acceptance.
• Required SDK5.1 native hardening, reviewed current-source policy, Q/E integration, exact-main acceptance and stable SDK publication remain dependencies. Legacy publication reports/readers are not overwritten to hide standalone freshness/count/navigation failures.
• This task has explicit user authorization for both own GitHub releases and their automatic Zenodo archival after all required gates pass. It does not answer or alter the separate approval, hooks, policies or review gates. No new release/tag/deposition has been performed.

AGEINT 0.2 Source and workflow cutover

Current Source: f98b9bb7701efcb4b5e470ed3bc36749b2a1878d. Comprehensive source/compiler, reader navigation, evidence/citation provenance and blocking native release workflows. Historical 0.1 readers and identifiers are not relabeled as current acceptance.

Exercised evidence

• Original clean-source 42383495664 passed 474 tests at 90.31% coverage, produced 333 reader files with 3,440 supported headings, and a real 1,826-page PDF. All non-task readiness checks passed; the aggregate correctly refused the unfinished SDK release dependency. Original inspection/diagnostic artifacts remain inspection-only.
• Actual inspection found a numbered link to unnumbered References rendered as an unresolved Pandoc-crossref placeholder despite the old green audit. Current code emits a named anchor link and rejects remaining HTML/TeX placeholders, including tables/headings. The new regression failed before the fix.
• Fresh current-code strict build and 33 reference/heading cases passed. Real Pandoc-crossref combined all 333 inputs; the native reference audit reports zero issues. Chromium clicked the actual named References link to its unnumbered target, loaded all 12 original bounded-decoder font faces and 177 figures with no broken images. This standalone navigation smoke explicitly used CDN mode with external browser requests blocked; it is not offline acceptance.
• Fresh complete native PDF/HTML/coverage/all-audit verification of this exact Source is 42399385517, using explicit SDK development candidate 71f3b52bf0bef7f2395766367bad0e69c53d6d34. A development candidate never grants release authorization.

Publication still gated

Genuine SDK v5.1.0 publication and authenticated assets; current stable-SDK full reader/PDF/offline/copied-code review; current task/readiness evidence and genuine production artifact promotion; ordinary main integration; exact-main same-run create-only v0.2.0 release custody.

Canonical legacy report bytes and existing reader/PDF artifacts are preserved pending genuine replacement. Text CC BY 4.0, code Apache-2.0 and the work-level DOI remain intact. No approvals, processes, locks or policies are overridden. No new archive deposition or package-index publication is authorized by this PR.

[docxology]

Original CI run 42360459120 has been retained and diagnosed, not rerun or waived. Python 3.11/3.12/3.15 each report 433 passed, 10 skipped and 84.96% coverage. All three artifact gates remain failing against the historical checked-in output tree: the new required output/data/source_citation_coverage.json is absent, freshness is false, and artifact evidence rejects that stale state. The release lane also correctly refuses the not-yet-published SDK v5.1.0 (HTTP 404). The independent exact-candidate native reader run 42360516743 continues; genuine production must replace the old artifacts before these gates can pass.

[docxology]

The original full native run 42360516743 reached 90.34% coverage with 448 passing tests, then correctly failed three contracts: the resolver-isolation test inherited the explicitly configured parent SDK; independent raw counts omitted the newly derived 564 authored-note field; and the public build used a private working namespace in 72 genuine diagnostic paths. No coverage floor or privacy/artifact gate is being weakened. The fixture now clears all supported explicit resolver bindings, the independent counter recounts raw note fields, and hosted public rendering is moving cleanly to the already supported active/AGEINT context. A separate review found valid longer/unclosed Markdown fences could be altered; new real regressions failed before the parser correction. Current native qualification is running. There is no successful inspection/release handoff from the failed run.