Provenance Record

Well-structured News-category piece (728 words, within the 400-1200 range) with clear Overview / What We Know / What We Don't Know sections. Attribution is dense and explicit ('according to X') throughout, and every direct quote is set off in quotation marks with a named speaker.

Read all 4 fetched snapshots from disk (sources/2026-08/kyoto-university-team-names-limnospondylus-ajimuensis-a-new-giant-salamander-genus-from-35-million-year-old-japanese-fossils/): source-1.html.gz (Sci.News), source-2.html.gz (ScienceDaily), source-3.html.gz (Phys.org, archive.org fallback), and source-0.html.gz (EurekAlert). source-0.html.gz turned out to be a 2179-byte AWS WAF JS challenge page, not real content, despite the fetcher recording status_code 202 (not flagged as unreachable by the automated checklist, but functionally unusable). peerj.com/articles/21362/ snapshot failed entirely (file: null, HTTP 403); I confirmed this is a genuine bot block, not a fetcher bug, by independently re-attempting both WebFetch (403) and a direct curl fetch with a browser user-agent (also 403) against the live PeerJ URL. Because the article's heaviest-cited source (EurekAlert, 10 attributions) was unreadable in the committed snapshot, I went further than a normal review: I fetched the live EurekAlert URL via curl (HTTP 200, 105KB of real HTML, saved to scratchpad for this review) and cross-checked it verbatim against every claim/quote attributed to 'Kyoto University's release' in the article body. All matched character-for-character, including both direct quotes from Masahiro Noda, the 3 June 2026 PeerJ publication date, the Tsubusugawa Formation spelling, the 1.1-meter size estimate, the three funders, the anterior-trunk/mid-trunk/sacro-caudal vertebra description, and the elephants-and-crocodiles co-fossil detail. All claims attributed to Sci.News (micro-CT and geometric morphometric methodology, the 2.6-million-year cooling onset, lake-vs-river habitat contrast, broader Ajimu ecosystem list) were verified in source-1.html.gz. All claims attributed to ScienceDaily (the 'five genera previously described' figure, which the article reasonably paraphrases as implying a sixth genus — ScienceDaily says 'making the recognition of another genus an important addition', not literally 'sixth genus', so this is a fair inference clearly not presented as a quote) were verified in source-2.html.gz. PeerJ itself is cited in the body exactly once, for the journal name/link and the 3 June 2026 date only ('The findings appeared in the journal [PeerJ](...) on 3 June 2026') — no specific technical claim in the article is sourced to PeerJ alone. The bibliographic facts normally sourced to PeerJ (paper title 'A new genus of giant salamander (Urodela, Cryptobranchidae) from the Pliocene of Japan', DOI 10.7717/peerj.21362, volume 14/article e21362, author Masahiro Noda et al.) are independently corroborated identically across Sci.News, ScienceDaily, and Phys.org.

No hallucinated quotes, no fabricated specifics, no misattributions found. Every number, name, date, and direct quote traces to a cited, human-readable source (three via committed snapshot, one via a documented live-fetch fallback for a source whose committed snapshot failed).

High-quality, thoroughly-attributed submission. The automated script's APPROVE_WITH_CORRECTIONS suggestion was driven entirely by source-reachability warnings (an unlisted domain, a bot-blocked fetch), not by any content inaccuracy — manual verification, including an out-of-band live fetch to compensate for a failed committed snapshot, found the article's claims, quotes, names, dates, and the flagged formation-name spelling to all be accurate and properly attributed. Verdict overridden to APPROVE; no corrections record filed because none is warranted.