Content Quality: Well-structured News piece (Overview / What We Know / What We Don't Know / Analysis). Word count 708, within the News 400-1200 range. Neutral, factual tone; appropriately hedges in 'What We Don't Know' that the primary Science paper itself was not directly accessible.
Source Verification: Read all snapshots from sources/2026-08/ancient-dna-from-two-chilean-mummies-provides-first-genetic-proof-smallpox-arrived-from-europe/ by decompressing each source-N.html.gz. source-0.html.gz (ScienceAlert) confirms verbatim: 1492-1631 CE date range, CAM9 lineage naming, the full Nakagome quotes ('The CAM9 genomes serve as a snapshot...', 'Recovering ancient viral DNA is extremely challenging...remarkable moment', 'We are working with DNA that is in low concentration...'), and the de la Fuente Castro attribution. source-2.html.gz (Popular Archaeology) confirms verbatim: 'nearly identical...same outbreak or by the same circulating strain', the CAM9 lineage description, and 'three to four million people, particularly among Indigenous populations with no prior immunity' — an exact match to the article's death-toll sentence. It also carries the citation footer confirming the underlying paper is 'The genomic identity of early smallpox in South America, Science, 30-Jul-2026, 10.1126/science.aee6957' — journal Science, not Science Advances. source-3.html.gz (Euronews) confirms verbatim: the de la Fuente Castro 'literally written into biological evidence' quote, the Patricia Foster 'Nobody would ever question...' quote, the 13-bone-samples detail, the leg-bone-fragment / 18-35-years-old detail, and the 'medieval European strains' genetic-position claim. source-4.html.gz (Cyprus Mail) confirms verbatim: 'Bruno Romero González, a geneticist and a doctoral student at Trinity College Dublin' (full name, matching the article, not just 'Bruno González' as abbreviated in Popular Archaeology), the Hispaniola/1518 quote, the Pizarro/1530s detail, 'Camarones Cove', the 30-to-35/18-to-20 ages, and the camelid-wool/bundle-burial quote. source-5.html.gz (PopSci) confirms the study's core finding and Nakagome/Romero González quotes via the page's embedded ArticleBody JSON-LD field. source-1.html.gz (Discover Magazine) failed to capture real content — despite a non-null filename and HTTP 202, the saved snapshot is an AWS WAF bot-challenge page (2.4 KB, 'window.awsWafCookieDomainList', 'JavaScript is disabled... we need to verify that you're not a robot'), not the article. Per the skill's fallback rule I WebFetched the live Discover Magazine URL as a last resort: it returned quotes for 'Camarones 9, a site used during the Inca and early colonial periods' and 'researchers screened bone samples from 13 people at Camarones 9. Only the woman and young man tested positive.' I treat this WebFetch result as corroborating rather than sole evidence, since the 13-sample/2-positive detail is independently confirmed verbatim in the Euronews snapshot itself. science.org (source-6, the primary DOI page) returned HTTP 403 (bot-blocked) with no snapshot saved — expected for this publisher; the journal/DOI/title were independently cross-verified via the Popular Archaeology citation footer instead, so the primary-source gap does not leave the DOI/title claim unverified.
Factual Accuracy: Independently verified all four items the contributor bot's PR notes said it excluded as likely WebFetch hallucinations, and confirmed none appear in the final article body (grep of body_markdown: 'Lara Cassidy'/'Cassidy' absent, '99.9' absent, 'Science Advances' absent, '90%'/'90 percent' absent). Went further and checked whether the underlying exclusion reasoning held up against the actual snapshots: (1) 'Science Advances' — confirmed false; the real journal is Science per both the Popular Archaeology citation footer ('...Science, 30-Jul-2026. 10.1126/science.aee6957') and Cyprus Mail ('published on Thursday in the journal Science'). The article correctly states 'journal *Science*' and cites the matching DOI. (2) 'Lara Cassidy' co-author quote — this one was NOT actually a hallucination. It appears verbatim in source-5.html.gz (PopSci, one of the submission's own 7 cited sources): '"...possibly aided by the lack of immunity newly exposed populations had," explained microbiologist and study co-author Lara Cassidy.' The bot's caution here was unnecessary (the detail was corroborated by a source already in its own source list), but since the published article never mentions Cassidy or attributes anything to her, this is a missed-detail omission, not a factual error — nothing false is stated. No correction needed. (3) '99.9% identical genome' figure — confirmed absent from every readable snapshot; a raw grep hit for '99.9' in the Cyprus Mail file turned out to be an SVG path-data coordinate, a false positive, not article text. Exclusion was correct. (4) '90% of Indigenous populations killed' — this figure DOES appear in Cyprus Mail ('smallpox and other viral diseases killed, according to some estimates, roughly 90% of the indigenous people'), so it is not a hallucination, but it is a single-sourced, more dramatic figure that conflicts in specificity with the two-source-corroborated 'three to four million' death toll the article actually uses (Popular Archaeology + Discover Magazine coverage, itself independently confirmed verbatim in Popular Archaeology). Preferring the better-corroborated figure over the single-sourced one was sound editorial judgment, not an error. Additionally verified character-for-character: 'CAM9' lineage name (ScienceAlert, Popular Archaeology) and 'Camarones 9' site name (Discover Magazine live fetch, cross-corroborated by the Euronews 13-sample detail) both match the article's usage exactly. All other specifics — dates (1492-1631), ages (30-35 / 18-20), burial details, Hispaniola/1518, Pizarro/1530s, and every direct quote in the body — trace verbatim to their cited sources. No hallucinated quotes, no unsourced specifics, no bidirectional source-array gap (all 7 submission.article.sources URLs are cited in the body and vice versa).
Overall Assessment: Clean, well-sourced News article. The contributor bot's own self-exclusion notes were independently checked against the raw source snapshots: two of the four flagged items (Science Advances, 99.9% figure) were correctly identified as unsupported/false and correctly excluded; the other two (Lara Cassidy, 90% figure) turned out to be real content in cited sources that the bot excluded out of excess caution rather than because they were hallucinated — but since neither appears in the published article, this is an omission, not a factual error, and requires no correction. All quotes, names, dates, and the CAM9/Camarones 9 naming verified character-for-character. Both automated 'warning' findings are administrative (allowlist config gap; one bot-blocked primary source with the DOI/journal independently cross-verified elsewhere) rather than content defects, so verdict is APPROVE rather than APPROVE_WITH_CORRECTIONS.