Verify a decision

{"actor":"judge","decided_at":"2026-08-26T08:03:27.669Z","decision":"review_revise","investigation_id":"4f465022-8404-4177-bf02-a31c259a5e60","new_score":0,"page_slug":"kelpdao-bridge-exploit-april-2026","prev_score":2,"reason":"Of 19 checked claims, 11 were confirmed, 2 partially supported, 1 disputed, 4 unverifiable, and 1 stale, putting disputed_pct at 26.3% (claim_findings[4] and the four unverifiable entries) — within the minor-issues band. The page's central facts (loss amount, attack mechanism, Aave exposure figures, and the Arbitrum freeze) are all confirmed and, notably, more accurate and internally consistent than two sibling AVOID.NET pages covering the same incident. The one substantiated problem is claim_findings[4]: the page implies the KelpDAO and Drift Protocol hacks trace to the 'same' DPRK subunit, while the cited analytical source (TRM Labs) treats them as distinct clusters — a real but secondary attribution overreach, not a core-fact error. The four unverifiable items (claim_findings[8], [16], [17], [18]) concern peripheral figures — a TVL percentage, a token-supply ratio, a protocol count, and product-naming details — that the review could not independently corroborate but also did not find contradicted, consistent with search-scope limits rather than page defects. claim_findings[11] is a completeness gap, not an inflation of wrongdoing: the page leaves the LayerZero/KelpDAO blame dispute open when LayerZero's later admission of fault actually resolved it in KelpDAO's favor. No coverage gap was flagged high priority, and reviewer confidence (0.72) is moderate, supporting a light penalty near the low end of the applicable band.","score_delta":-8,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}